Mutations that impair Eyes absent tyrosine phosphatase activity in vitro reduce robustness of retinal determination gene network output in Drosophila

A limited collection of signaling networks and transcriptional effectors directs the full spectrum of cellular behaviors that comprise development. One mechanism to diversify regulatory potential is to combine multiple biochemical activities into the same protein. Exemplifying this principle of modularity, Eyes absent (Eya), originally identified as a transcriptional co-activator within the retinal determination gene network (RDGN), also harbors tyrosine and threonine phosphatase activities. Although mounting evidence argues for the importance of Eya’s phosphatase activities to mammalian biology, genetic rescue experiments in Drosophila have shown that the tyrosine phosphatase function is dispensable for normal development. In this study, we repeated these rescue experiments in genetically sensitized backgrounds in which the dose of one or more RDGN factor was reduced. Heterozygosity for sine oculis or dachshund, both core RDGN members, compromised the ability of phosphatase-dead eya, but not of the control wild type eya transgene, to rescue the retinal defects and reduced viability associated with eya loss. We speculate that Eya’s tyrosine phosphatase activity, although non-essential, confers robustness to RDGN output.


Introduction
Only tens of signaling pathways and roughly one thousand transcription factors operate in metazoan species [1][2][3]. One means of increasing regulatory potential is to incorporate feedback that dampens or amplifies signaling flux depending on context [4]. Another strategy is to evolve modular proteins with a physically separable set of biochemical capabilities that can be harnessed in different combinations appropriate to specific developmental, cellular or subcellular situations [5].
Less understood are Eya's two protein phosphatase activities: a threonine phosphatase domain sandwiched between two portions of the N-terminal TAD [21,22] and a tyrosine phosphatase whose key catalytic residues are dispersed in linear sequence across the ED [23][24][25]. Focusing on the tyrosine phosphatase, initial overexpression experiments in Drosophila suggested functional relevance [23,24,26], but a subsequent loss-of-function based analysis concluded that Eya's tyrosine phosphatase activity is not required for normal Drosophila development [27]. Specifically, using fully functional genomic BAC rescue transgenes, Jin et al. showed that introducing missense mutations biochemically proven to disrupt tyrosine phosphatase function in vitro and in mammalian cultured cells did not compromise development, fertility, or survival of the fly [27]. In light of the identification of physiologically relevant substrates for mammalian Eya tyrosine phosphatases [28][29][30][31], together with the high degree of evolutionary conservation of the residues that form the phosphatase catalytic core, we found the dispensability of Eya's tyrosine phosphatase activity in the fly puzzling [32].
Here, we tested the alternative model that Eya's tyrosine phosphatase function confers robustness to RDGN regulatory output by re-examining the genetic rescue capability of the phosphatase-dead eya BAC transgene under conditions of genetic stress. Sensitizing the network with eya, so and/or dac heterozygosity, neither of which on its own or in combination caused overt phenotypes, revealed that "phosphatase-dead" eya lacked sufficient activity to support normal retinal development and survival. While the most parsimonious interpretation is that Eya's tyrosine phosphatase activity contributes to overall RDGN output, until physiologically relevant substrates are identified, an equally plausible possibility is that the "phosphatase-dead" mutations disrupt some other function or interaction. We hope our work will motivate further study of Eya's possible enzymatic activities in Drosophila.
For the genetic rescue experiments, crosses were set with 2-3 adult flies of each sex, allowed to lay eggs for 3-4 days, transferred to a fresh vial for another 3-4 days before discarding the parents. Adult progeny were counted daily until all had eclosed. Survival was measured by comparing the expected Mendelian ratio of rescued straight winged eya cli ;GR/+ animals relative to their curly winged (eya cli /CyO,act-GFP;GR/+) siblings. Crosses were performed at least five times in parallel.
To assess female fertility, rescued female flies (eya cli ;GR/+) were collected as virgins, fed yeast for 2d at 25˚C, and then crossed in pairs to three w 1118 males. Parents were tossed after 9 days and progeny were counted daily until all flies eclosed. Ten crosses were set in parallel, but only those in which both females survived were scored.
To image adult eyes, 3-5 day old adult flies were decapitated and photographed with a Canon EOS Rebel camera fitted to a Leica dissecting microscope. Individual slices were merged using iSolution-Lite software (IMT-Digital).

Adult retina embedding and sectioning
Adult heads were decapitated, halved, fixed for 30 min in 2.5% glutaraldehyde + 1% O S O 4 in 0.1 M sodium phosphate buffer pH 7.2 for 30 min on ice, and incubated in 2% O S O 4 in 0.1 M sodium phosphate buffer pH 7.2 for 2h on ice. Retinas were then dehydrated with successive 10 min incubations on ice in 30%, 50%, 70%, 90%, and twice 100% EtOH. The tissue was incubated 3x 10 min on ice in propylene oxide, placed in 50% propylene oxide/50% soft Durcupan resin (Sigma) for 12h at room temperature, and then in 100% resin for 4h at room temperature. Retinas were then transferred to BEEM flat embedding molds (Ted Pella) and baked at 70˚C for 12h, cut into 2 μm sections with a microtome, and mounted in DPX (Sigma). Samples were imaged at 100x under phase contrast.

Results and discussion
Lowering so and dac levels reveals a requirement for Eya's tyrosine phosphatase activity during retinal development Under physiologically wild type conditions, Eya's tyrosine phosphatase activity is dispensable for development and survival [27]. In contrast, overexpression-based genetic rescue and ectopic eye induction assays have found that the same phosphatase-dead mutations diminish or ablate function, suggesting the phosphatase can function in Drosophila cells [23,24,26]. This discrepancy led us to hypothesize that there might be sufficient regulatory redundancy within the RDGN to compensate for loss of phosphatase activity under physiological conditions but not under overexpression conditions. If so, then lowering the dosage of other RDGN factors, which on its own does not compromise normal development or survival, might sensitize the system to reveal the essential contributions of the phosphatase. We selected So and Dac to test this idea because of their intimate genetic and physical relationship with Eya in the RDGN [6,[35][36][37][38], The three genomic rescue (GR) transgenes used in our experiments, eya + GR, eya E728Q GR and eya D493N GR, were generous gifts from Graeme Mardon and colleagues [27].
We first assayed Eya's tyrosine phosphatase contribution to retinal development by rescuing eya loss with GR transgenes in genetically sensitized backgrounds and analyzing adult eye morphology. In the control experiment, heterozygosity for so or dac or for both so and dac did not alter the ability of one copy of the eya + GR wild type rescue transgene to support normal eye morphology in an eya cli null background (Fig 1, first two columns). Flies rescued with two copies of any of the three GR transgenes were indistinguishable from wild type (Fig 1A) in these heterozygous backgrounds. In contrast, in all three sensitized backgrounds, the external morphology of eyes from flies rescued with one copy of the phosphatase-dead eya E728Q GR or eya D493N GR transgenes was no longer fully wild type (Fig 1, last four columns). Across the board, qualitative assessment of these fully penetrant rough eye phenotypes indicated that eya E728Q GR rescued eya loss less efficiently than eya D493N GR. Phenotypically, in the eyes of the former, the regularity of the ommatidial lattice was noticeably disrupted when viewed under a dissecting microscope, and upon sectioning, some photoreceptor loss was apparent. In contrast, eyes rescued by eya D493N GR exhibited mild disruption of the ommatidial lattice, particularly near the posterior margin and less frequent photoreceptor loss. Based on qualitative assessment of the adult rough eye phenotypes, for both eya E728Q GR and eya D493N GR, reducing the dose of so sensitized the system more effectively than reducing the dose of dac (Fig 1, compare middle two rows), while the so, dac double heterozygote heightened sensitivity to loss of phosphatase activity (Fig 1, bottom row). To rule out the possibility that the phenotypes reflected interactions with secondary mutations on the eya cli chromosome, we repeated all of the crosses shown in Fig 1 in different transheterozygous backgrounds, including the same eya cli /Df(eya) used by Jin et al., 2013, eya cli /eya G130 and eya cli /eya A188 , and observed identical results (Fig 1). We propose that when Eya, So and Dac are limiting, Eya's tyrosine phosphatase activity may be required to achieve sufficient RDGN output to support normal retinal development.
To pinpoint more precisely when Eya's tyrosine phosphatase is needed for RDGN function, we compared the pattern of the differentiating photoreceptor neurons in the third instar eye fields of eya + GR versus eya E728Q GR rescued larvae (Fig 2). In agreement with the phenotypes observed in adult animals (Fig 1), qualitative assessment of the field of Elav+ cells in discs from larvae carrying one copy of eya E728Q in a background of reduced so, dac or so and dac dosage showed that it was under-developed compared to wild type (compare Fig 2D, 2F, 2H to 2C, 2E and 2G). Ommatidial organization was also irregular, consistent with the adult rough eye phenotype. Eya protein levels and expression pattern appeared qualitatively similar across the genotypes (Fig 2), as previously reported [27]. We conclude that when other RD proteins are limiting, one copy of Eya E728Q lacks sufficient activity to support normal patterning within the developing eye field.
In the course of these experiments, we were surprised to find that even in animals carrying the normal dose of so and dac, eya E728Q GR was insufficient for normal adult eye formation. Thus in contrast to the published study [27], in our hands, animals carrying one copy of the eya E728Q GR transgene but null for endogenous eya developed eyes that were not completely wild type (Figs 1A-1D and 2B). We observed a qualitatively similar mild and fully penetrant phenotype in eya cli homozygotes and in eya cli /Df(eya) trans-heterozygotes (Fig 1C), indicating the defects were not caused by homozygosed secondary mutations on the eya cli chromosome. Consistent with the prior report [27], the eyes of animals heterozygous for eya D493N GR were indistinguishable from wild type (Fig 1E and 1F).
Given the different results obtained with the two phosphatase dead GR transgenes, we worried the eya E728Q GR chromosome might have accumulated a secondary mutation. We therefore outcrossed to w 1118 flies for ten generations and then repeated the analysis. The results matched our initial observations (Fig 1C,1I,1O and 1U). Finally, we sequenced all protein coding exons from the two phosphatase dead BAC transgenes. This analysis confirmed the presence of the expected E728Q and D493N missense changes and the absence of other mutations that would alter the protein sequence. However it remains formally possible that a secondary lesion elsewhere in the eya E728Q GR BAC or in a locus closely linked to the transgene insertion site is responsible for the phenotypes observed in our rescue experiments.
While the molecular explanation for the activity differences between eya E728Q GR and eya D493N GR remains unclear, analogous differences have been noted in other contexts. For example, in in vitro phosphatase assays using recombinant murine Eya3, although both the D493N and E728Q equivalent mutations significantly reduced catalytic activity, the former still had measurable phosphatase activity while the latter did not [23]. Greater loss of functionality in eya E728Q than eya D493N has also been reported in overexpression-based ectopic eye induction and genetic rescue assays [23,24]. Assuming the two mutations reduce phosphatase activity to different extents in vivo as they do in vitro, then the phenotypic distinctions between eya E728Q GR and eya D493N GR may reflect a threshold of sensitivity in the requirement for phosphatase activity. Alternatively, the different "phosphatase-dead" missense mutations might disrupt different protein-protein interactions or subtly perturb Eya stability, subcellular localization or other functions in ways our analyses were unable to detect. For example, variations in Eya-So transcriptional functions and interactions could explain the activity differences we noted between eya E728Q GR and eya D493N GR. Considering the overexpression studies in cultured cells and in the fly that show that "phosphatase-dead" missense mutations do not abrogate So-binding or broadly alter Eya-So transcriptional output [20,23], we favor the hypothesis that the "phosphatase-dead" mutants actually impair catalytic activity in fly cells as they do in vitro and in mammalian cells, and that in genetically sensitized backgrounds, i.e. those in which Eya, So and Dac levels are half normal, this perturbs development. However until a substrate is identified, this caveat must be kept in mind.
Eya's tyrosine phosphatase may contribute to the RDGN's role in survival but is dispensable for fertility eya, so and dac are considered essential genes, though dac null flies occasionally reach adulthood [6,36,[38][39][40][41], raising the possibility that Eya's tyrosine phosphatase is also required for survival. To evaluate this, we crossed eya cli /CyO; eyaGR to eya cli /CyO and then asked how closely the ratio of eya cli ; eyaGR/+ to eya cli /CyO; eyaGR/+ adult progeny matched the 1:2 Mendelian ratio expected under the null hypothesis that the eyaGR transgene fully rescues the lethality associated with complete eya loss. eya E728Q GR reduced the proportion of homozygous eya cli animals that eclosed from the expected 0.3 ratio to 0.2 (Fig 3A), suggesting that Eya's tyrosine phosphatase contributes to biological processes important for viability. so or so,dac heterozygosity exacerbated this phenotype, whereas dac heterozygosity partially suppressed. In contrast, animals carrying one copy of eya D493N GR eclosed at the expected rate, even in the sensitized backgrounds. We confirmed that these results were independent of second chromosome background by repeating several of the experiments with eya null transheterozygotes or null alleles over a deficiency that covers the eya locus ( Fig 3B). These data indicate that Eya's tyrosine phosphatase may cooperate with normal if they had the wild type complement of rhabdomeres and abnormal if they did not. Statistical significance was assessed with a Fisher's exact test; * indicates P < 0.05. https://doi.org/10.1371/journal.pone.0187546.g001 other RD proteins in contexts outside the eye and that "phosphatase-dead" mutants can compromise viability when these proteins are limiting. We also assessed the role of Eya's phosphatase in female fertility by counting the number of progeny produced by pairs of eya cli ; eyaGR/+ females outcrossed to wild type males. In contrast to our measurements of retinal development and viability, and in agreement with the earlier study [27], females rescued with the phosphatase dead BAC transgenes were as fertile as controls, including when so and/or dac levels were reduced (Fig 4). eya cli ; eyaGR/+ males were also fertile in all genetic backgrounds, although we did not quantify or compare fertility rates.

Concluding remarks
Our work raises the possibility that Eya's tyrosine phosphatase activity contributes to the robustness of RDGN output during normal retinal development, analogous to its requirement for efficient ectopic eye induction. If correct, this implies that regulation of retinal specification in the eye disc carries sufficient redundancy that it is insensitive to modest genetic stress, including two-fold changes in RDGN protein levels and loss of Eya phosphatase function; only if multiple such perturbations are introduced simultaneously is this buffering capacity overwhelmed. In contrast, we propose that the requirement for higher than usual RDGN output to drive ectopic eye induction makes the system inherently less robust. Thus, while the overexpression-based ectopic eye induction assay may not provide an accurate estimate of RDGN regulatory redundancy during normal retinal development, its inherent lack of robustness makes it exquisitely sensitive, facilitating the discovery of relevant biological regulation. We therefore suggest that the debate as to whether or not Drosophila Eya participates in the RDGN as a protein tyrosine phosphatase should remain open and hope our study will motivate further investigations, including a search for physiological substrates.