In Drosophila, early eye development is governed by a set of DNA binding proteins and transcriptional coactivators that collectively are termed the retinal determination (RD) network. The core factors are encoded by the Pax6 genes eyeless (ey, Quiring et al.,1994; and twin of eyeless [toy], Czerny et al.,1999), the SIX class transcription factor sine oculis (so, Cheyette et al.,1994; Serikaku and O'Tousa,1994), the transcriptional coactivator and protein tyrosine phosphatase eyes absent (eya, Bonini et al.,1993) and a distant relative of the Ski/Sno protooncogene dachshund (dac, Mardon et al.,1994). Removal of any of these genes leads to the complete elimination of the adult compound eye (reviewed in Kumar,2008). Remarkably, forced expression of any member within nonretinal tissues such as the antennae, legs, halters, and wings can lead to the formation of ectopic eyes (Halder et al.,1995; Bonini et al.,1997; Shen and Mardon,1997; Czerny et al.,1999; Weasner et al.,2007). As a cassette, these factors function during eye development in a broad range of organisms including vertebrates (reviewed in Treisman,1999; Wawersik and Maas,2000; Hanson,2001). In addition to this core network of genes, there are several other factors that interact with and modulate the activities of these genes, which are also required for early eye development or can induce ectopic eye formation. These genes include the Pax6(5a) genes eyegone (eyg, Jun et al.,1998) and twin of eyegone (toe, Jang et al.,2003), the SIX transcription factor optix (Seimiya and Gehring,2000), the Meis1 homolog homothorax (hth, Pai et al.,1997), the serine-threonine kinase nemo (nmo, Braid and Verheyen,2008), the pipsqueak motif containing DNA binding proteins distal antenna (dan) and distal antenna related (danr, Curtiss et al.,2007) and the zinc finger transcription factor teashirt (tsh, Pan and Rubin,1998).
Tsh was first initially identified as a specifier of trunk identity and segmentation in the embryo, where loss of tsh function resulted in the transformation of the prothoracic segment to a labial identity (Fasano et al.,1991; Roder et al.,1992). A potential role for tsh during early eye development was first hinted at when it was recovered in an enhancer/promoter (EP) screen for genes that could force nonretinal tissues into adopting an eye fate (Pan and Rubin,1998). Tsh protein is distributed throughout the regions of the developing eye that lie anterior to the advancing morphogenetic furrow (MF). It functions to promote cell proliferation and depending upon the context can either promote or inhibit eye development (Bessa et al.,2002; Singh et al.,2002,2004; Bessa and Casares,2005). Retinal mosaic clones that completely lack tsh develop normally thereby prompting the suggestion that a second gene is acting in a functionally redundant manner to tsh in the eye (Pan and Rubin,1998).
The tsh paralog, tiptop (tio) was identified as a regulator of Drosophila embryogenesis (Laugier et al.,2005). tio null mutants are viable, fertile and display no retinal abnormalities. tsh and tio have different expression patterns until stage 17 of embryonic development, after which point they have shared expression domains. Interestingly, these genes are not completely redundant to each other as knockdown experiments with RNAi constructs can induce retinal defects. It appears that both genes are maintained, in part, because each gene is capable of repressing its own expression and that of its paralog (Laugier et al.,2005; Bessa et al.,2009). Thus, it seems that correct development of the retina is not dependent upon the absolute expression levels of each individual gene. Rather, it is suggested that the eye simply requires certain combined levels of this zinc finger subfamily (Bessa et al.,2009).
In this study, we have attempted to further elucidate the roles of Tsh and Tio in eye development. We find that both genes are expressed in identical patterns in the eye–antennal disc and that their expression levels are not significantly different from each other. We have also analyzed the ability of the duplicates to direct eye specification and demonstrate that while tsh and tio can induce ectopic eye formation (Pan and Rubin,1998; Bessa et al.,2009), this activity is limited in that both genes are capable of only redirecting the fate of the developing antennal imaginal disc. Additionally, we observe that tio is more effective than tsh in inducing ectopic eye formation.
tsh and tio Are Expressed at Similar Levels in the Eye
Tsh/Tio proteins belong to a subclass of Zn finger DNA binding proteins. Tsh contains three such motifs while Tio, like its orthologs in basal insects and vertebrate species, contains four domains (Fig. 1A; Fasano et al.,1991; Laugier et al.,2005). Both tio and tsh are expressed in coincident patterns ahead of the advancing morphogenetic furrow (Fig. 1B,C; Pan and Rubin,1998; Bessa et al.,2009). tio null mutants and tsh null retinal clones do not appear to adversely affect eye development (Pan and Rubin,1998; Laugier et al.,2005). This has prompted the suggestion that each gene can substitute for the other during retinal development. In support of this hypothesis, is the demonstration that each gene negatively regulates the transcription of the other. The loss of either gene would result in the de-repression of the other paralog, thus compensating for the initial deficit in expression of either tsh or tio (Laugier et al.,2005; Bessa et al.,2009). One expectation from this model is that the transcriptional levels of both genes should be approximately equal. To test this hypothesis, we used quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) to determine and compare the transcriptional levels of both tsh and tio in the developing eye disc. We recovered RNA from eye–antennal imaginal discs and observe that a typical wild-type eye–antennal disc contains approximately 8,356 tsh and 7,787 tio transcripts, respectively. These differences (across five samples and with multiple primer pairs) are not statistically significant (P = 0.627). We conclude that both genes are not expressed at significantly different levels from each other (Fig. 1D). These results further support the premise that tsh and tio are likely to share many common functions during normal eye development.
Differences in Ectopic Eye Formation Are Not Due to Expression Patterns or Levels
Despite multiple similarities, these genes are unlikely to be completely functionally redundant as tio can only partially rescue the trunk denticle phenotype of tsh null mutants (Laugier et al.,2005). As retinal development in tio null mutants and tsh null mutant clones is relatively normal we set out to determine whether there are any differences between the abilities of either Tsh and Tio to induce ectopic eye development. Previous reports have demonstrated that both genes can induce ectopic eyes when driven by dpp-GAL4 (Pan and Rubin,1998; Bessa et al.,2009). We extended these results by forcibly expressing tsh and tio (individually) throughout development with 219 different GAL4 drivers and observe that both genes can coax only cells within the antennal disc into adopting a retinal fate (see below). Coexpression of both genes did not lead to an expansion of tissues that can be converted the retinal tissue—ectopic eyes are still restricted to the antennal disc. We found that expression of tsh could induce ectopic eyes with four GAL4 drivers (Fig. 2A–D) while tio could induce ectopic eyes when expressed with eight drivers (Fig. 2A–H). To attribute this result to an actual property of the gene, we tested (1) temporal and spatial differences in the driver expression pattern; (2) the transcriptional level of each UAS line; and/or (3) the transcriptional strength of each GAL4 line. We first compared the expression patterns of the drivers and do not see any correlation between the pattern of expression and the induction of ectopic eyes. For example, tsh can induce ectopic eyes when expressed with cb41-GAL4 but not with MJ33a-GAL4 even though they are expressed in very similar patterns within the antennal disc (Fig. 2C,G). All eight drivers are expressed in both the second and third-instar discs; thus, it does not appear that temporal cues are critical.
We then measured the levels of GAL4 transcription using quantitative RT-PCR. Although we observe significant differences between individual driver lines, there is no correlation between these differences and the ability of each gene to induce ectopic eyes (Fig. 2I,J). Among the four drivers that produce ectopic eyes when placed in combination with UAS-tsh (dpp-GAL4, cb26-GAL4, cb41-GAL4, and Ser-GAL4) the average transcript level of GAL4 ranges from 212 to 14,825, suggesting that the transcription level of the these particular driver lines is not the limiting factor in the induction of ectopic eyes by tsh. In addition, expression of tsh with dpp-GAL4, cb26-GAL4, or cb41-GAL4 induces ectopic eyes but not with T98-GAL4 even though the average transcript level of the latter (6,089) is substantially greater than those of the former lines (Fig. 2I,J). We then turned our attention to the expression levels of the two UAS lines themselves. Individual expression of tsh and tio in all cells posterior to the furrow by means of the GMR-GAL4 driver induces a mild roughening of the posterior portions of the adult retina (Fig. 3A,C). The retinal phenotypes are nearly identical to each other. Using quantitative RT-PCR we measured the transcript levels in GMR-GAL4/UAS-tsh and GMR-GAL4/UAS-tio third-instar imaginal discs (Fig. 3B,D). These measurements include not only the transcripts that are derived from the UAS line but also the endogenous tsh and tio transcript levels respectively. We then subtracted the average wild-type endogenous tsh and tio transcripts and determined an average transcript level that should only reflect transcription from the UAS lines themselves. We observe that GMR-GAL4/UAS-tsh and GMR-GAL4-UAS-tio eye imaginal discs produce (directly from the UAS insertion) approximately 1,782 tsh and 3,330 tio transcripts, respectively (Fig. 3E). These differences (across four biological samples and with multiple primer pairs) are not statistically significant (P = 0.126), and we conclude that, in this case, the transcriptional levels of the UAS lines themselves do not account for the differences in ectopic eye formation.
Tsh Induces Ectopic Eyes in Two Cell Populations
We then focused on the location of the ectopic eyes within the developing antennal segment. Within the developing antennal segment two zones can be coaxed into adopting a retinal fate. The first and most commonly transformed territory is the ventral-most region of the antennal disc (Fig. 4A–D, 6; green arrow). This area normally gives rise to the head cuticle that lies adjacent to the eye disc (Haynie and Bryant,1986). It should be noted that most other retinal determination genes are capable of inducing ectopic eyes only within this zone (Halder et al.,1995; Bonini et al.,1997; Shen and Mardon,1997; Seimiya and Gehring,2000; Weasner et al.,2007; Braid and Verheyen,2008). The second, less frequently converted region is within the portion of the antennal disc that will give rise to the arista and possibly the third antennal segment of the adult antenna (Fig. 4A, green arrowhead; Haynie and Bryant,1986). Other than tsh and tio, the only other genes that are known to convert these cells into photoreceptor cells are distal antenna (dan) and distal antenna related (danr; Curtiss et al.,2007).
The ectopic eyes, particularly those derived from along the ventral ridge of the antennal disc, can range in size, with some eyes containing just a handful of ommatidia (Fig. 4D,H) to others being almost half the size of the normal compound eye (Fig. 4C,G). The larger eyes appear to be the result of significant tissue proliferation as the antennal discs are greater in size than wild-type and have an abnormal profile. This is consistent with the role that has been assigned to tsh in promoting cell proliferation in the normal eye (Bessa et al.,2002). It also appears that there is a distinct directionality to movement of the ectopic furrows. Instead of seeing furrows all along the internal edge of the ectopic eye, as is seen in patched (ptc) and Pka loss-of-function clones as well as hedgehog (hh) overexpressing clones (Chanut and Heberlein,1995; Heberlein et al.,1995; Ma and Moses,1995; Pan and Rubin,1995; Strutt et al.,1995) we observe with antibodies directed against DAC and EYA proteins that an ectopic furrow is present only along a partial stretch of the internal edge (Fig. 4E,F, white arrows).
Tio Is a More Effective Inducer of Ectopic Eyes
Expression of tio appears to transform the same two antennal cell populations into ectopic eyes (Fig. 5A–G, green arrows and arrowheads). Similar to those induced by tsh, the ectopic eyes that result from tio expression can also range in size from just a few ommatidia (Fig. 5D,E,G,K) to nearly half the size of the normal eye (Fig. 5H,J,K). Expression of tio also appears to promote tissue proliferation as the antennal discs appear distorted and significantly larger in size than wild-type (Fig. 5C,F–K). However, despite these similarities there are also several differences between the two genes. First, in addition to the four GAL4 drivers that induced ectopic eyes with tsh (Figs. 2A–D, 5A–D), tio induced ectopic eyes when expressed with four additional GAL4 drivers (Figs. 2E–H, 5E–G). Of interest, when tio is expressed with one of these four drivers (T98-GAL4), we fail to observe ectopic eyes in third-instar imaginal discs but do see them in adult animals (data not shown). It is likely that the final steps of eye development (i.e., cell fate specification) are executed during the early stages of pupal development. Second, unlike tsh, expression of tio often induces multiple ectopic eyes within a single antennal disc (Fig. 5E,I,K). And third, we observe that, in many instances that an ectopic furrow forms along the entire internal edge of the ectopic retina (Fig. 5H, arrow). Simultaneous expression of both tsh and tio did not increase the number of GAL4 drivers that could induce ectopic eyes. In fact, ectopic retinal development remained limited to the eight GAL4 drivers described in Figure 2.
In an attempt to identify paralog-specific genetic interactions that might explain the distinctions between tio and tsh, we conducted modifying genetic screens on flies that were expressing either tsh or tio in all developing photoreceptors by means of the GMR-GAL4 driver. Expression of either gene leads to a near identical mild to moderate roughening of the posterior portion of the retina (Fig. 3A,C). Underlying this roughening is a nonspecific loss of photoreceptor cells within each ommatidium (data not shown). We conducted two genetic screens, identified several interacting genes including wg and homothorax (hth) but failed to identify any paralog-specific interactions. The identification of genetic interactions with hth are consistent with recent demonstrations that expression of tio, like that of tsh, activates and maintains hth expression (Bessa et al.,2009). Our failure to identify such interactions does not rule out their existence, but it does suggest that such interactions are not exceedingly common and might make only a minor contribution to their subtle functional differences.
In this report, we describe several findings that shed light on the function of tsh and tio, two zinc finger containing transcription factors that appear to function during eye development in Drosophila. Our first result focuses on the transcript levels of both genes during normal retinal formation and we find that both paralogs are expressed at equivalent levels. Previous reports have demonstrated that loss-of-function tio null mutants and loss-of-function tsh null retinal clones have negligible retinal defects (Pan and Rubin,1998: Laugier et al.,2005). Both genes are expressed in identical patterns anterior to the furrow (Bessa et al.,2009) and have the ability to repress their own and each other's expression (Laugier et al.,2005; Bessa et al.,2009). From these data, a model has emerged in which tsh and tio share responsibilities and are partially redundant in the eye, where the loss of one gene leads to the transcriptional de-repression of the paralog. The result is compensation by one paralog for the loss of the other. In the simplest incarnation of this model, both genes should be expressed at relatively equivalent levels. If either gene is expressed at significantly higher levels than its paralog, then mutations in the lesser-expressed gene should result in a visible phenotype. For example, the paralogs eyegone (eyg) and twin of eyegone (toe) are expressed in identical patterns during eye development. However, their transcript levels vary, with eyg and toe transcripts make up 87% and 13% of the total Pax6(5a) transcript pool. Mutations in eyg have drastically reduced eyes, whereas knockdown of toe expression by RNAi has minimal if any phenotypic effects (Yao et al.,2008). Having both genes (tsh and tio) expressed at similar levels supports the observations that loss of either gene has minimal phenotypic effects and that both genes repress each other's transcription without extinguishing each other's expression.
The second result is focused on the ability of tsh and tio to induce ectopic eye development in comparison with other RD genes. Prior reports have demonstrated that, when expressed in the dpp expression pattern, both genes are capable of inducing ectopic eye formation (Pan and Rubin,1998; Bessa et al.,2009). We used 219 unique GAL4 drivers to express UAS-tsh and UAS-tio responder lines and find that the ability of both tsh and tio to induce ectopic eye formation differs significantly from ey, which can promote eye development in a plethora of nonretinal tissues including the legs, antennae, wings, and halteres (Halder et al.,1995). In contrast, ectopic eye formation, induced by the expression of tsh and tio, is limited to just the developing antennal segment. Another difference between tsh, tio, and the other eye specification genes is in the nonretinal cell populations that they can convert into eye tissue. Unlike other eye specification genes, which can only transform a single cell population within the antenna (Halder et al.,1995; Bonini et al.,1997; Shen and Mardon,1997; Seimiya and Gehring,2000; Weasner et al.,2007; Braid and Verheyen,2008), tsh and tio can coax two separate cell populations into adopting a retinal fate (Fig. 6). These results suggest a developmental paradox. On one hand, tsh and tio appear more effective than other eye specification genes including ey in converting cells within the antenna toward a retinal fate. On the other hand, however, both genes are much more restricted than genes like ey and eya in inducing ectopic eyes in other nonretinal tissues.
Finally, we focused much of our efforts toward understanding the extent to which the two genes differ functionally. We have observed that, while tio can induce ectopic eyes when expressed with eight different GAL4 drivers, tsh can only induce ectopic eyes when expressed with a subset of these drivers. Several observations suggest that tio is a more effective inducer of ectopic eyes than tsh. First, tio can induced ectopic eyes when expressed through a wider range of GAL4 drivers. Second, tio appears capable of inducing multiple ectopic eyes within a single antennal disc while tsh does not. And third, the ectopic eyes that are induced by tsh appear to have a fully formed morphogenetic furrow while only partial furrows are generated by the expression of tsh. There did not appear to be any correlation with these observations and the spatial and temporal expression patterns of the GAL4 drivers or with the transcriptional levels of either the GAL4 drivers or the UAS responder lines. We also attempted, but ultimately failed to identify paralog-specific genetic interactions that could explain this phenomenon.
The appearance of Tsh and Tio appears to be a lineage-specific duplication event, occurring before the Drosophilid diversification. It is a distinct possibility that these paralogs have subfunctionalized in Drosophila, although polymorphism data are required to confirm the speculation (Shippy et al,2008; Bessa et al., 2009). Studying these genes in the context of eye development not only reveals new players in the RD cascade, it unearths a new regulatory feedback loop during imaginal disc development. We can also show what paralogs are doing in the nascent stages of gene evolution by studying subtle functional changes like those we have described here.
Despite the coincident expression of tsh and tio and the lack of retinal phenotypes in individual null mutants, there is a singular reason to suspect that these proteins have at least a subset of distinct functions: Tio and Tsh are structurally distinct with Tsh harboring three zinc fingers DNA binding regions while Tio has four (Fasano et al.,1991; Laugier et al.,2005). A tsh/tio gene with three zinc fingers appears to be specific to the Drosophilids as homologs within basal insects and vertebrate species all contain four putative DNA binding domains (Caubit et al.,2000,2005; Laugier et al.,2005; Shippy et al.,2008; R.R. Datta and J.P. Kumar, unpublished data). Furthermore, there are also differences in the nonconserved segments of the proteins. The differences related to the induction of ectopic eyes are likely due to disparities (1) in the specificity and/or affinity of DNA binding; (2) in the number of zinc finger domains; or (3) in the nonconserved regions of the protein. Identifying, this difference is likely to shed considerable light onto how the Tsh and Tio proteins function during normal and ectopic eye development and will also show whether coding subfunctionalization is occurring along the lengths of the paralogs.
The following fly stocks were used in this study: UAS-tsh (gift of Albert Courey), UAS-tio (gift of Laurent Fasano), GMR-GAL4 (gift of Larry Zipursky), tio-GAL4 (gift of Amit Singh), UAS-ey (gift of Walter Gehring), UAS-GFP and the DrosDel Deficiency and GAL4 collections (gifts of the Bloomington Drosophila Stock Center). All GAL4 crosses were conducted at 25°C.
The following primary antibodies were used in this study: rat anti-ELAV (1:100, DSHB), mouse anti-DAC (1:5, DSHB), mouse anti-EYA (1:5, DSHB), rabbit anti-TSH (1:3,000, Stephen Cohen). The following secondary antibodies were obtained from Jackson Laboratories: donkey anti-mouse tetrarhodamine isothiocyanate (TRITC; 1:100), donkey anti-rat fluorescein isothiocyanate (FITC; 1:100), goat anti-rat FITC (1:100), goat anti-rat TRITC (1:100), goat anti-rabbit (1:100). F-actin was detected with phalloidin-TRITC (Molecular Probes). Imaginal discs were dissected in phosphate buffer, fixed in 4% paraformaldehyde, washed in wash buffer (0.1% Triton), and then incubated in primary overnight. Secondary antibody incubations lasted 2–3 hr after which tissues were further dissected in wash buffer and then mounted onto slides in Vectashield (Vector Laboratories). Tissues were examined using a Zeiss Axioplan 2 compound microscope with Apotome and then imaged using a Zeiss Axiocam MRm camera. Adult flies with ectopic eyes were frozen at −80°C for 20 min, imaged using a Zeiss Discovery light microscope and photographed with a MRc color camera.
RNA from the eye–antennal discs of third-instar larvae from the following stocks were isolated using the RNeasy Micro Kit (Qiagen): w1118, GMR-GAL4/UAS-tsh, GMR-GAL4, UAS-tio, dpp-GAL4, cb26-GAL4, cb41-GAL4, Ser-GAL4, c329b-GAL4, C833-GAL4, MJ33a-GAL4, T98-GAL4. The Sybr Green Two Step RT-PCR Kit (Invitrogen) was used for reverse transcription, cDNA synthesis and quantification. A total of 1 μg of total RNA was used as the starting template for each reaction. The samples were quantified using the Stratagene MxPro3000P qPCR system.
RT-PCR Primer Sequences
The following primer pair was used to detect GAL4 transcripts: 5′-TTCTTCGTCGACGATGC-3′ and 5′-AATTGGTTAGAGCGGTG-3′. The following primer pairs were used to detect tsh transcripts: (primer set 1) 5′-TCCGCGAGCTGAGACGAAAAGAG-3′ and 5′-CGGGGCGAAGGCAAGGCG-3′, (primer set 2) 5′-TCCGCGAGCTGAGACGAAAAGAA-3′ and 5′-CGGGGCGAAGGCAAGGCG-3′, (primer set 3) 5′-TCTGTAGGTACCCGGAAACG-3′ and 5′-TTCCAGTCAGGGAATTGACC-3′, (primer set 4) 5′-AGGAATCTTCAAAGCCAGCA-3′ and 5′-TGGCACTTCCATTTACCACA-3′. The following primer pairs were used to detect tio transcripts: (primer set 1) 5′-TGGGTCACAGATTGCAGACACG-3′ and 5′-GTTAAACAGTCGGCTTCGTAAA-3′, (primer set 2) 5′-TTGGGTCACAGATTGCAGACACG-3′ and 5′-GTTAAACAGTCGGCTTCGTAAA-3′, (primer set 3) 5′-GACAAAGCTTCCGGTCTCTG-3′ and 5′-GACGGAACTCCAGTGTTGGT-3′, (primer set 4) 5′-AGGAATCTTCAAAGCCAGCA-3′ and 5′-TGGCACTTCCATTTACCACA-3′.
Gene Interaction and Ectopic Eye Screens
GMR-GAL4/UAS-tsh and GMR-GAL4/UAS-tio flies were crossed to the 356 stocks that constitute the DrosDel Deficiency Kit. F1 progeny were scored for modifications of the rough eye phenotype that results from expressing either tsh or tip in the developing eye. We then identified putative interacting genes by crossing the above screen stocks to single gene disruption mutations (that are located between the breakpoints of modifying deficiencies) and scoring for the similar modifications as seen with the overlying larger deficiency. UAS-tsh and UAS-tio were crossed to 219 GAL4 drivers and adults were scored for the presence of ectopic eye formation. All crosses were conducted at 25°C.
We thank Amit Singh, Laurent Fasano, Albert Courey, Larry Zipursky, and Walter Gehring for gifts of fly stocks. J.P.K. was funded by the National Institutes of Health.