Single base pair differences in a shared motif determine differential Rhodopsin expression

Abstract

The final identity and functional properties of a neuron are specified by terminal differentiation genes, which are controlled by specific motifs in compact regulatory regions. To determine how these sequences integrate inputs from transcription factors that specify cell types, we compared the regulatory mechanism of Drosophila Rhodopsin genes that are expressed in subsets of photoreceptors to that of phototransduction genes that are expressed broadly, in all photoreceptors. Both sets of genes share an 11bp activator motif. Broadly expressed genes contain a palindromic version that mediates expression in all photoreceptors. In contrast, each Rhodopsin exhibits unique single bp substitutions that break the symmetry of the palindrome and generate activator or repressor motifs critical for restricting expression to photoreceptor subsets. Novel sensory neuron subtypes can therefore evolve through single base pair changes in short regulatory motifs, allowing the discrimination of a wide spectrum of stimuli.

In the visual system, different photoreceptor neurons express specific light-sensing pigments (1); however, common downstream factors amplify and convert the response to the visual stimulus into a neuronal signal. For instance, each unit eye (ommatidium) of the Drosophila retina contains eight photoreceptors (R1-R8) that express different light-sensing Rhodopsins (Rhs) that are restricted to specific photoreceptor subsets. Outer photoreceptors R1-R6 express Rh1. Inner photoreceptors R7/R8 express either Rh3 in pR7s coupled with Rh5 in pR8s, or Rh4 in yR7s with Rh6 in yR8s (Fig. 1A) (1). R1-R8 all share broadly expressed phototransduction factors (Fig. 1B and fig. S1A) that amplify and convert the response to the visual stimulus into a neuronal signal (2).

Fig. 1. Broadly expressed and restricted photoreceptor genes share a cis-regulatory motif.

(A) In p ommatidia, Rh3 in pR7 is coupled with Rh5 in pR8, whereas in y ommatidia, Rh4 in yR7 is coupled with Rh6 in yR8. Outer photoreceptors R1-R6 express Rh1. Right: Cross-section at the level of R7 (top) or R8 (bottom).

(B) Broadly expressed photoreceptor genes (7 upper genes) and restricted Rhodopsins Rh1-Rh6 share the 11 base pair P3/RCSI motif in their proximal promoters.

Left: All motifs contain a 5’ TAAT homeodomain core binding site, which is repeated in reverse orientation (ATTA) in broadly expressed phototransduction genes. The 3’ ATTA is modified in RCSI motifs of Rh1-Rh5. In Rh6 and Rh3, central bp differences (orange) create K₅₀ sites (TAATCC) for the activator Otd and the repressor Dve (9). Q₅₀ sites (TAATTG/A) are bound by the photoreceptor-specific activator Pph13 (6).

Right: Reporter expression patterns of the broadly expressed phototransduction gene trpl and the restricted Rh6 at the R8 level. Retinas were stained for GFP (green), Rh5 (blue) and Rh6 (red). Scale bars, 10 µm.

Bottom: Pph13 and Otd are expressed in all photoreceptors, whereas Dve is expressed at high levels in R1-R6 and at low levels in yR7s.

Here, we examine the cis-regulatory mechanisms that distinguish restricted from broad expression patterns for Rhodopsins and downstream phototransduction factors, respectively. All Rhs share the conserved Rhodopsin Core Sequence I/RCSI (3, 4), which resembles the palindromic P3 motif (TAATYNRATTA), an optimal binding site for Paired-class homeodomain proteins (5). Almost all known broadly expressed phototransduction genes contain a P3 motif in their proximal promoter (Fig. 1B, fig. S1A, and Supplementary Text). The presence of a conserved P3/RCSI motif within 100bps of the Rh transcription start site (TSS) is significantly associated with enrichment in adult eyes (χ-squared test, p-value < 0.001). P3/RCSI is required for activation in photoreceptors since its mutation caused either a loss or a strong reduction in expression of 16 broad or restricted reporters (figs. S1, S2 and S3), with the exception of Arr1 (fig. S2K). Moreover, expression of 10 out of 15 reporters was lost in mutants for the photoreceptor-specific transcription factor Pph13 (Fig. 1B, figs. S2 and S3), a Paired-class homeodomain protein that binds P3 and the Rh6 RCSI in vitro (6, 7).

Since each Rh promoter has a highly conserved RCSI variant (Fig. 1B) (4), we tested the sufficiency of P3 and RCSI to determine the significance of the specific differences between perfectly palindromic (P3) and imperfect motifs (RCSI) (Fig. 2). Four copies of the P3 motif (including four neighboring bps for spacing; the contribution of these additional bps was only tested for Rh4, see below) from the broadly expressed ninaC, rdgA, or trpl, drove broad expression in all photoreceptors (Fig. 2, A and A’, and fig. S4, A and A’), consistent with our previous results (8). In sharp contrast, multimerized RCSI motifs drove expression in subsets of photoreceptors. The RCSI of Rh3 and Rh6 contains a K₅₀ motif, a binding site for K₅₀ homeodomain proteins such as the Dve repressor or the Otd activator (Fig. 1B). Expression of [Rh3 RCSI]₄ and [Rh6 RCSI]₄ was biased to inner photoreceptors: [Rh3 RCSI]₄ mediated restricted expression in R8 and in R7, with a strong bias towards the pR7 subset where Rh3 is normally expressed (Fig. 2, B and B’). This pattern is complementary to the expression of Dve (Fig. 1B) (9), which is indeed responsible for the restricted expression as [Rh3 RCSI]₄ drove a broad, P3-like pattern in dve mutants (Fig. 2 B’’). [Rh6 RCSI]₄ drove restricted expression in R8s and R7s; expression in R1-R6 was very weak in comparison to P3 motifs, which was due to dve-dependent repression (fig. S4D, D’ and D’’).

Fig. 2. The P3 motif is sufficient to drive expression in all photoreceptors, while RCSI motifs drive expression in subsets of photoreceptors.

Multimerization of a P3 motif (left) found in broadly expressed phototransduction genes or of a specific RCSI motif (right) from a restricted Rhodopsin.

(A, A’) Tetramerization of a palindromic P3 motif from ninaC results in broad, pan-photoreceptor expression.

(B, B’, B’’) Four copies of the Rh3 RCSI that contains a Dve repressor site (orange) drive expression in R7 and R8. (B) The reporter is strongly biased towards pR7s (arrows), where Rh3 is expressed, and faint in yR7s. (B’) GFP is expressed in all R8s, which lack Dve. (B’’) Expression is expanded to all photoreceptors in a dve¹⁸⁶ mutant background.

(C, C’, C’’) Tetramerization of the Rh1 RCSI drives variable reporter expression (arrows and arrowheads in C’’) in individual R1-R6 photoreceptors, where Rh1 is expressed (blue in C’’). Scale bars, 10 µm.

(D) RCSI motifs are biased towards the respective endogenous Rh expression pattern (‘wild type’).

[Rh1 RCSI]₄ drove variable expression in R1-R6 (Fig. 2, C and C’), where Rh1 is expressed. This outer photoreceptor-specific pattern is complementary to the inner photoreceptor expression of [Rh3 RCSI]₄. Rh4 has the same RCSI as Rh1. However, adding the synergistic 3’ RCSII motif (fig. S1, F–I) (3) led to expression in yR7s, where Rh4 is expressed (fig. S4, B and B’, and fig. S4C and C’). Although [Rh5 RCSI]₄ was not sufficient for reporter expression (fig. S4, E and E’), adding three K₅₀ motifs to a single Rh5 RCSI ([K₅₀]₃+[Rh5 RCSI]₁) led to expression in R8 and pR7 photoreceptors (fig. S4, F and F’).

In summary, the RCSI motifs of specific Rhs differ from palindromic P3 motifs in broadly expressed genes: They drive expression that is biased towards the endogenous Rh expression patterns (Fig. 2D). We show below that full subtype-specificity and activation often requires the repetition of motifs that are present in the RCSI.

As specific RCSI motifs directed restricted expression in different photoreceptor subsets (Fig. 2D), albeit with incomplete subtype-specificity and with some variability in expression levels, we asked whether the single bp differences are required for subtype-specificity in a wild type promoter context, and which other motifs are required for full restriction. We mutated the K₅₀ (Otd/Dve) motifs (TAATCC) to Q₅₀ (Pph13) motifs (TAATTG/A) (Fig. 1B) to disrupt repression while preserving RCSI-mediated activation. Mutating the Rh3 RCSI resulted in an expansion to yR7s where Dve is present at low levels (Fig. 3, A and B). Mutating the Rh6 RCSI caused de-repression in R1-R6 and the ocelli (fig. S5, A and B, and fig. S6A). Rh3 and Rh6 have K₅₀/Dve repressor motifs repeated upstream and mutation of individual motifs also caused de-repression in yR7s (Fig. 3C) (10) and R1-R6/ocelli (fig. S5D and fig. S6B), respectively. Taken together, single bp changes create K₅₀ motifs in the Rh3 and Rh6 RCSI, which are required for subtype-specific expression together with their upstream repeats.

Fig. 3. Single base pairs in the RCSI motifs of Rh3 and Rh4 are critical for photoreceptor subtype-specific expression.

(A) The Rh3 reporter is specifically expressed in pR7s. Left: Rh3 promoter containing an upstream ATTC/yR7 repressor motif (yR7R), two K₅₀ motifs and the Rh3 RCSI composed of a K₅₀ motif and an ATTC/yR7 repressor motif. (B–F) Mutation of the Rh3 RCSI (B, D, E) or of its partial upstream repeats (C, F) causes de-repression in yR7s.

(G) The Rh4 reporter is specifically expressed in yR7s. Left: Rh4 promoter with the Rh4 RCSI and RCSII motif (3). (H, I) Mutating a single bp in the Rh4 RCSI causes de-repression in pR7 and R1-R6 that depends on Pph13. Scale bars, 10 µm.

(J) De-repression in other photoreceptor subsets (indicated by ‘X’) caused by mutations of RCSIs or upstream repeats.

We also examined the significance of the disrupted P3 palindrome, i.e. the imperfect 3’ homeodomain binding motif in the RCSIs of Rh1-Rh5 (Fig. 1B). Creating a palindromic motif in the Rh3 RCSI (TAATCCAATTC→TAATCCAATTA) caused de-repression in yR7s (Fig. 3D) that depended on Pph13 (Fig. 3E). Therefore, de-repression appears to be due to increased activation through the newly created Q₅₀/Pph13 site. The same ATTC→ATTA mutation in the Rh5 RCSI led to partial de-repression in yR8s (fig. S5, E and F). This single bp change created a binding site for the activator Otd (AGATTA) (11) and indeed de-repression in yR8s was lost in otd mutants, as was activation in pR8s (fig. S5G).

The 3’ ATTC motif in the RCSI of Rh3 and Rh5 is repeated upstream. Mutating the upstream repeat without creating a Q₅₀/Pph13 site (ATTC→CAAA) also caused de-repression in yR7s (Rh3) or yR8s (Rh5) (Fig. 3F and fig. S5H). Mutating both ATTCs of Rh5 enhanced de-repression into almost all yR8s (fig. S5I). Therefore, we have identified repressor motifs in the RCSIs of four Rhs (K₅₀/Dve motifs in Rh3/Rh6 and ATTC motifs in Rh3/Rh5). These motifs are repeated upstream within less than 100 bps and are required for full subtype-specificity.

A single bp ATTT→ATTA mutation in the Rh4 RCSI caused de-repression in R1-R6, pR7s and the ocelli (Fig. 3, G and H, and fig. S6D). The correct pattern was restored by crossing the mutant Rh4 reporter in a Pph13 mutant background (Fig. 3I), indicating that the A→T change prevents Pph13 from overcoming repression in the ‘wrong’ photoreceptor subsets, as was the case for Rh3 and Rh6. The same mutation in the Rh1 RCSI caused no detectable de-repression (fig. S5J). Replacing two bps in the RCSI of the ocelli-specific Rh2 (fig. S6, E and F) to obtain a Q₅₀/Pph13 site led to de-repression in R1-R6 photoreceptors that depended on Pph13 (fig. S6, G and H).

Our in vivo data reveal that a cell-fate decision requires single bp differences in RCSI motifs (Fig. 3J). They complement previous findings in cell culture that subtle sequence differences in a glucocorticoid receptor or NF-κB binding site can specify the mode of transcriptional regulation (12, 13), and that small differences in binding site sequences can lead to distinct Hox specificities in vivo and in vitro (14–16). (i) Single bps in RCSI prevent binding of dimers of broadly expressed activators such as Pph13 (5) (fig. S8B), tipping the balance of activator/repressor binding. This weakened activation allows repressors to prevent activation in other photoreceptor subtypes (17). (ii) They generate specific combinations of overlapping activator and repressor motifs, often repeated upstream to provide robust expression and full subtype-specificity. Creating overlap of activator and repressor motifs is an efficient way of blocking a key activator site in the ‘wrong’ cell types that express a repressor (18), especially since the RCSI motifs are very close to the transcription start site and repression there could block other activators (19). The precise tuning of RCSI motifs within their respective promoter context leads to incompatibility in other Rh promoters, as revealed by RCSI swap experiments: Replacing a given RCSI with another one resulted in two main outcomes, loss of expression or de-repression in specific subsets of photoreceptors (fig. S7 and Supplementary Text).

The RCSI/P3 motif resembles ‘terminal selector’ motifs that allow the coordinated expression of effector genes that define a particular neuron type (20, 21). Yet, RCSI motifs exhibit additional layers of regulation that are integrated in a single regulatory element, as their sequence is modified for subtype-specificity. Mutating a cis-regulatory motif in many cases appears to be the shortest evolutionary path towards a novel phenotype (22). Although we found that it is possible in some cases to eliminate ectopic expression by removing the broadly expressed activator Pph13 (Fig. 3, E and I), this simultaneously causes a loss of expression of several broad phototransduction genes, defects in photoreceptor morphology and a severe loss of light sensitivity (23).

We propose that the modification of a P3-type motif into different RCSI-type motifs allowed partitioning Rh expression to different subtypes of photoreceptors (Fig. 4 and fig. S8). This opened the possibility to discriminate wavelengths and likely conveyed a selective advantage. In this model, P3 motifs represent a positive regulatory element shared by ancestral genes that were expressed in all photoreceptors. This regulation is conserved, as the promoter of the long wavelength Rh as well as Gβ76C that are both expressed in all photoreceptors in the beetle Tribolium, contains a palindromic P3-type motif and depend on Pph13 (24, 25).

Fig. 4. Modification of a shared cis-regulatory motif for color vision.

(A) A palindromic P3 motif (TAATNNNATTA) provides broad activation of an ancestral Rh and a set of phototransduction (PT) genes in all photoreceptors (left). Modification of single bps yields an RCSI motif (orange) that is essential for restricting Rh expression to subsets of photoreceptors (right). The upstream repetition of parts of the RCSI (orange box) is required for full subtype-specificity. Right schematic: phototransduction cascade. Note that the downstream acting factors (broad PT genes) remain expressed in all photoreceptors.

(B) Palindromic P3 motifs bound by a photoreceptor-specific Q₅₀ activator like Pph13 provide broad activation, while single bp changes in RCSI motifs specific to each Rh create novel activator or repressor motifs (right).

In conclusion, our study revealed a high level of precision at every base pair in a short cis-regulatory element that is critical for proper spatial (broad or restricted) expression. It will be interesting to see whether similar modifications of shared cis-regulatory motifs are used to diversify neuronal cell types in other developmental contexts, for instance in human photoreceptor and olfactory genes (fig. S9, Supplementary Text and Tables).