The Eyes Absent proteins in development and in developmental disorders

The Eyes Absent (EYA) proteins are part of the well-conserved retinal determination gene network (RDGN) that contributes to cell-fate determination processes in multiple tissues and species (reviewed in [1–4]). In addition to the EYA proteins, the RDGN consists of the master regulatory transcription factor PAX6 (eyeless ey and twin of eyeless toy in flies), the SIX family of homeodomain transcription factors (SIX1–6; sine oculis and optix in flies), and the transcription factor DACH (dachshund). Interactions between components of the RDGN are summarized in Figure 1. While the RDGN was first described in Drosophila eye development [5], surveys of the eukaryote genome suggest that the RDGN was employed in two different contexts early in evolution, eye development and oocyte development [6]. In vertebrates, the proteins coded by members of the RDGN are widely implicated in cell survival and differentiation during tissue specification. Here, we describe our current understanding of the roles played by EYA proteins during development and in tissue homeostasis. Particular emphasis is placed on linking the multiple biochemical activities of the EYA proteins with their developmental roles whenever possible. We will highlight how the EYA proteins interface with other signaling pathways involved in organ development, the emerging evidence that EYA3 contributes to photoperiodism, and mutations in the Eya genes associated with human disorders.

Mammals have four Eya paralogs (Eya1–4) and their expression begins during embryonic development. Eya expression is much more restricted to specific adult tissues, with Eya3 having the most wide-spread expression. The intrinsically cytoplasmic EYA proteins physically partner with the SIX proteins and are then translocated to the nucleus where they play an essential role in the regulation of gene expression driven by the RDGN (Figure 1b).

All animal EYA proteins are characterized by a highly conserved C-terminal ∼270 amino-acid domain referred to as the EYA domain (ED) (Figure 1a). Interaction with the SIX proteins occurs through the ED and there is strong evidence that ED has tyrosine phosphatase activity [7–9]. However, ED bears no structural or sequence similarity to the classical protein tyrosine phosphatases (PTP). This places the EYA proteins in the unique position of being PTPs that do not employ the nucleophilic Cysteine residue-mediated catalytic mechanism characteristic of the classical PTPs (recently reviewed in [3]). Few EYA-PTP substrates have yet been identified (recently reviewed in [3]). The best-characterized of these is the DNA damage repair-associated histone protein H2AX [10,11]. Dephosphorylation of the terminal C-tyrosine of H2AX promotes DNA damage repair and survival of cells. In addition, the tumor suppressor estrogen receptor-β and the cytoskeleton modulating protein WDR1 have also been identified as likely EYA-PTP substrates [12,13]. The PTP activity of the EYA proteins also promotes cell motility possibly via the Cdc42/Rac GTPases [14,15], but no EYA substrate has yet been identified in this context. The N-terminal domain of the EYA proteins has an independent transactivation function localized to a Pro/Ser/Thr-rich ∼210 amino-acid segment in Drosophila Eya [16]. In addition, a threonine phosphatase activity (either intrinsic or via interaction with protein phosphatase 2A (PP2A)) is associated with the N-terminal domain [17–22].

This juxtaposition of distinct biochemical activities in a single polypeptide places the EYA protein in the unique position of participating directly in both the regulation of gene transcription and in signal transduction processes. A schematic overview of molecular pathways impacted by EYA proteins as they participate in the processes described below is provided in Figure 2.

Loss of the eya gene in Drosophila eye progenitor cells anterior to the furrow (eya² mutant flies) resulted in viable but completely eyeless flies due to apoptosis of eye progenitor cells [23]. The introduction of mouse Eya2 in progenitor cells prior to furrow formation in the eya² mutant background was able to restore eye formation, speaking to a high degree of conservation of function for these EYA proteins through evolution [24]. Initial reports suggested that MAPK-catalyzed phosphorylation of Eya positively regulates fly eye development [25–27]. However, subsequent investigation revealed that MAPK-mediated Eya phosphorylation was dispensable for normal eye development or survival [28].

Whether the tyrosine phosphatase activity of Drosophila Eya contributes to eye development has been addressed multiple times with varying conclusions. The original studies using overexpression or mis-expression showed that phosphatase-dead forms of Eya were defective in the rescue of the eya-null eyeless phenotype [8,9]. However, when genomic transgenes mimicking endogenous spatial and temporal levels of Eya were used in rescue experiments, phosphatase-dead Eya complemented the eya null alleles [29]. When such dose-controlled rescue experiments were conducted in genetically sensitized backgrounds (heterozygosity for other RDGN components sine oculis or dachshund), the phosphatase-dead form of Eya was not able to support normal retinal development [30]. Hence the current thinking is that Drosophila Eya is not essential on its own, but rather contributes to the ‘robustness’ of RDGN output in fly eye development [30].

It is not straightforward to extrapolate from fly eye development to the more complicated process of vertebrate eye development. In flies, development initiates at a single imaginal disc epithelium. In contrast, in vertebrates, the optic vesicle evaginates from the forebrain and forms the optic stalk, the retina and the retinal pigment epithelium, while the lens and cornea derive from the surface ectoderm. Eya1 is seen earliest in the ectoderm-derived lens placode, Eya2 in the retina, and Eya3 in the optic vesicle. While no gross eye phenotype has yet been reported upon knockout of any Eya gene in mice, there is one report of congenital cataracts and anterior segment anomalies in humans with an Eya1 mutation [31]. Eya1^−/− mice die at birth with severe cranio-facial and skeletal defects. They also lack thymus, parathyroid glands, ears and kidneys [32]. Except for open eyelids, there appear to be no eye development defects. Eya2^−/− and Eya3^−/− mice have no obvious phenotype and are viable and fertile [33,34]. No eye phenotype has been recorded in double Eya1^−/−/Eya2^−/− mice [33]. The only reported ocular effect upon loss of an EYA protein is a mild delay in the development of the post-natal mouse retinal vasculature upon endothelial-specific deletion of Eya3 [35].

In this context, it is relevant that the loss of other components of the RDGN have variable effects on vertebrate eye development. Six3-null mice lack all forebrain structures including eyes [36]. Loss of Dach1, which is expressed in the developing retina, results in no eye phenotype [37]. Pax6^−/− mice die at birth with a complete disruption of eye development [38], and heterozygous Pax6 mutation in humans results in small eyes [38]. Dosage effects of the Pax6 gene on eye formation have been demonstrated through the use of the CRISPR/Cas system to generate a mosaic mixture of mutant and wild-type mice [39]. Notably, lens development from the surface ectoderm requires higher Pax6 gene dosage than retinal development from the optic vesicle [39]. These observations suggest that the roles of the RDGN proteins in eye development are more complex than can be deciphered from simple loss-of-function experiments, and are likely to be influenced by genetic background, somatic mosaicism and perhaps environmental factors.

EYA1 contributes to multiple stages of inner ear development. The inner ear has six sensory organs and each one has sensory hair cells as well as non-sensory supporting cells. Eya1 and the transcription factor Sox2 are co-expressed in the otocyst where pro-sensory epithelia are formed and play a role in specifying the sensory cell lineage [40]. EYA1 along with SIX1 and the BRG1-associated SW1/SNF chromatin remodeling complex (BRG1-BAF) induce Sox2, and thereby promote the specification of a pro-neurosensory-restricted progenitor cell fate from an ectodermal cell [41]. The model proposed by these studies is that EYA1 bridges SIX1 and the BRG1-BAF complex, thus inducing the formation of a compact structure that enables binding to distant cis-regulatory elements, and hence Sox2 induction [41]. Together, EYA1 and SOX2 are required for the expression of sensory markers including Jag1, Lfng, Bmp4, and Fgfs, and for cell cycle progression of these pro-sensory progenitors. Once the progenitors become committed to a hair cell fate, Sox2 expression is no longer detected. However, Eya1 continues to be expressed in differentiating hair cells along with Atoh1. Consistent with this, Eya1-null mice completely lack inner ear sensory organs. There is evidence that dose-dependent EYA1 activity is necessary for the development of different parts of the ear [40], mirroring the variation in inner ear malformation seen in patients with Eya1 mutation-associated developmental disorders (described below). According to an Eya2^lacZ knock-in reporter study Eya2 is also expressed in differentiating hair cells in the sensory epithelia of the inner ear and Eya2^−/− mice have a mild hearing deficit [41]. These new studies once again point to the likelihood that Eya genes can act synergistically during development, as well the possibility that they can compensate for each other and complicate loss-of-function studies. No studies have yet specifically queried the involvement of the C-terminal PTP activity of the EYA proteins in ear development.

EYA proteins are implicated in various stages of trunk myogenesis. As in the case of the ear, the different Eyas are expressed at different timepoints during muscle development, and they work in conjunction with at least three SIX proteins. During early myocyte specification and myogenesis, Eya1 and Eya2 are expressed in the developing limb buds. In vitro reporter assays show that the N termini of these EYA proteins are sufficient to drive transactivation [42]. While some studies conclude that there is no requirement for EYA1 or SIX1 in early myogenesis [43] and that Eya1 knockout mice show no muscle phenotype [44], others report that Eya1 knockout mice have delayed myogenesis between e11 and e13, and that Eya1^−/−/Eya2^−/− mice have severe muscular abnormalities [33]. EYA1 also plays an important role in post-natal muscle development where it can switch a slow-twitch muscle to a fast-twitch phenotype by replacing the myosin I and IIA heavy chain isoforms to the fast-acting IIB and IIX isoforms [45]. Eya3 knockout mice are also reported to have weaker muscle strength [34].

RDGN member SIX1 also plays a critical role in muscle development; Six1 mutant mice have impaired myogenesis and extensive muscle defects in the forelimb, hindlimb and diaphragm [46]. There is a complete absence of all hypaxial muscles and a severe reduction in epaxial muscle in Six1^−/−/Eya1^−/− mice [44]. The EYA1-SIX1 complex activates SIX1 target genes involved in myogenesis including Pax3, MyoD and Myogenin [33]. Indeed, ectopic expression of Eya1 and Six1 (along with Esrrb (Estrogen related receptor-beta) and Pax3) is sufficient to induce myogenic stem cells with differentiation capacity in vitro and in vivo [47]. In flies, eya and so induce tinman expression (ortholog of mammalian NKX2-5), which in turn activates the JAK/STAT pathway to initiate myocyte differentiation [48]. In Caenorhabditis elegans, myogenic vs non-myogenic fate is determined by the SIX-EYA complex downstream of WNT signaling [49].

No specific examination of whether the tyrosine phosphatase function of the EYA proteins contributes to muscle development has been reported, and the implication in all of the studies cited above is that EYA acts as an activator of transcription in complex with the SIX proteins.

Eya1^+/− mice have renal abnormalities while Eya1^−/− mice completely lack kidneys [32]. Ureteric bud outgrowth and consequently metanephric induction are compromised in the absence of EYA1. Downstream of Eya1, Gdnf expression is down-regulated in Eya1-null mice. Eya1 is one of two genes (the other is Osr1) that are required for the initial formation of the metanephric mesenchyme which condenses to form a precursor cell population. These precursors can either differentiate (epithelialization to form the renal vesicle, the precursor of the nephron), or self-renew to maintain a progenitor pool at the tip of the ureteric bud (cap mesenchyme). A balance between self-renewal and differentiation is crucial to appropriate kidney development. Eya1 (and Six2) are expressed in the cap mesenchyme throughout nephrogenesis representing an essential multipotent cell population [50].

Nuclear EYA1 in nephron progenitors stabilizes MYC, which in turn promotes the proliferation of undifferentiated cells. This occurs through dephosphorylation of Thr58 on MYC, thus preventing MYC degradation and increasing MYC half-life. EYA1-mediated MYC stabilization induces nephron progenitor cell proliferation [50]. Threonine phosphatase activity in the N-terminal domain of the EYA proteins had been previously shown to play a role in the innate immune response [18]. While no structural or sequence similarity with known phosphatase families could be found in the poorly conserved N-terminal domain of the EYAs, threonine phosphatase activity was confirmed by several follow-up studies [17,19]. However, a recent report has raised the possibility that the threonine phosphatase activity ascribed to the EYA proteins is actually mediated through an interaction with the PP2A B55α subunit [21]. But in glioblastoma stem cells EYA1 directly interacts with MYC, and the EYA1-MYC axis regulates GSC proliferation, migration and self-renewal [51]. Regardless of the mechanism through which it occurs (either directly mediated by EYA or via PP2A), links between EYA proteins and MYC stabilization continue to emerge in contexts beyond kidney development (Figure 2).

Knockout studies show that the development of several neural-crest derived cranial structures are EYA1-dependent. Eya1^−/− embryos have defective pharyngeal arches and lack proximal arch structures contributing to a broad spectrum of defects in cranio-facial morphogenesis [52]. Interestingly, this is another developmental context in which a putative EYA1 threonine phosphatase activity is implicated: threonine dephosphorylation of the Notch intracellular domain (NICD) is believed to stabilize the NICD and thus enhance Notch signaling, a crucial contributor to cranio-facial development (Figure 2).

EYA1 is also required for the morphogenesis of organs derived from the third pharyngeal pouch (thymus, parathyroid, thyroid). Eya1^−/− mice display thyroid hypoplasia and the thymus and parathyroid fail to form [53]. Eya1^−/− mutants also have a spectrum of abnormalities in cardiovascular development, the most common being aortic arch defects [54]. This is exacerbated in Six1^−/−/Eya1^−/− double mutants (cardiac outflow tract and aortic arch defects), partially due to a reduction in fibroblast growth factor 8 (Fgf8) levels [54].

EYA1, along with SIX1, contributes to hindbrain development through its ability to promote Nrp1 transcription and favor Gli activators over Gli repressors following sonic hedgehog (Shh) stimulation [55]. In keeping with this role, Eya1 is expressed in Shh-subtype medulloblastomas and promotes tumor growth [55]. A Shh-Eya1 axis is also involved in promoting symmetric cell divisions during cerebellar development [22]. This is believed to occur through dephosphorylation of Thr410 and thus inactivation of the atypical protein kinase aPKCζ [22], once again implicating an EYA-linked threonine phosphatase activity in development.

A recent report shows that Eya1 is expressed in the bitter taste receptor cells of the oral epithelium and in the differentiating taste bud cells in circumvallate and fungiform papillae and the soft palate [56]. The signaling pathway(s) through which EYA contributes to the generation and differentiation of taste buds (specifically bitter cells) remains to be established.

An intriguing association between EYA3 and photoperiodism is emerging largely through studies on large seasonal animals and birds. Photoperiodism is a phenomenon through which functional and behavioral responses accommodate changes in the length of dark and light cycles, thus allowing plants and animals to adjust to seasonal variation. In animals, photoperiodism impacts phenomena such as growth, metabolism, migration, reproductive behavior, and hibernation. In mammals, light-sensitive retinal ganglion cells (RGCs) transmit light information to the suprachiasmatic nucleus (SCN) in the hypothalamus (Figure 3). This signal is then transmitted via light-sensitive neurons to the pineal gland, where nocturnal melatonin is synthesized. The length of the melatonin signal is consequently proportional to the length of the dark period, or night, and is read by the pars tuberalis (PT) in the pituitary gland [57]. Long days (summer) induce thryroid stimulating hormone (TSH; a heterodimer of α and β subunits) in the PT. TSH in turn promotes thyroid hormone production and activity in the basal hypothalamus, thus interfacing with the regulation of crucial season-guided physiology such as reproduction.

An EYA3-SIX1-TEF complex up-regulates TSHβ transcription in the PT [58]. EYA3 expression in the PT is rhythmic, peaking transiently 12.5 h after the onset of darkness or melatonin expression. As a result, in winter EYA3 peaks late in the night while in summer EYA3 peaks at dawn. Eya3 expression is thus central to the PT coincidence timing mechanism. Furthermore, through a positive feedback loop EYA3 promotes its own transcription and amplifies the EYA3-dependent coincidence-timer. As a consequence, EYA3 is a key player in the ability to switch between the winter and summer phenotypes that follow from gradual changes in day length in the Spring and Fall [59].

The tyrosine phosphatase-dead form of EYA3 was able to potentiate TEF/SIX1-mediated TSHβ transcription to levels comparable to the wild-type protein [58], suggesting that the tyrosine phosphatase activity is dispensable for EYA3's role in photoperiodism. Collectively, the currently available evidence supports a role for EYA3's transactivation activity in measuring night length and tuning the circadian clock with seasonal changes [60].

The importance of EYA proteins (and the RDGN network) during crucial stages of embryonic development is reflected in the numerous developmental disorders associated with Eya mutations. These are typically manifest as defects in eye, ear, neck, cranio-facial, and heart development (summarized in Table 1). Below we describe the developmental defects frequently associated with Eya mutations and, when available, links between the biochemical activities of the EYA proteins and these disease states.

Eya1 is one of the three known causative genes for BOR (Branchio-oto-renal syndrome also known as Melnick–Fraser syndrome; the other two genes being RDGN members SIX1 and SIX5 [61,62]). BOR is an autosomal dominant disorder characterized by malformations of the neck, ears and kidney. These include (to varying degrees) branchial fistulae or cysts, hearing loss as a result of inner, middle and outer-ear malformations, and kidney abnormalities ranging from hypoplasia to agenesis. Features of BOR partially overlap with otofaciocervical syndrome (OFC), a rare developmental defect characterized by distinctive facial features (triangular face, broad forehead, narrow nose and mandible, high arched palate), prominent ears, long neck, branchial fistulae or cysts, sloping shoulders and clavicles, hearing impairment and mild intellectual defects. Eya1 mutations are also associated with OFC [63]. Furthermore, Eya1 mutations are sometimes seen in patients with congenital cataracts and ocular segment anomalies, sometimes associated with BOR [31]. Deletions, mis-sense mutations as well as complex rearrangements of the Eya1 gene are reported in BOR and OFC patients [61,64–70]. Most of the mutations are unique to individual families and no apparent genotype–phenotype correlation has yet emerged. Several experimental studies tried to link individual mis-sense mutations to biochemical activities, variably reporting impaired tyrosine phosphatase activity as well as impaired interaction with the SIX proteins [71,72].

Eya1 mutations have been reported in patients with congenital cardiac abnormalities [73,74], and mutation of Eya4 is associated with dilated cardiomyopathy [75]. Genetic studies using compound deletions in mice show that EYA1 and SIX1 play a role in cardiovascular development, and it is proposed that a Tbx1-Six1/Eya1-Fgf8 regulatory cascade underlies a common mechanism important for the morphogenesis of both the heart and the face [54]. This data has led to the speculation that the pathogenesis of BOR and cardio-facial syndrome could be linked to Eya1 and Six1 mutations via a common molecular program [54].

Eya4 mutations are reported in several families with non-syndromic autosomal dominant post-lingual progressive hearing loss (late-onset deafness) [76–79]. Most of these mutations result in truncated EYA4 protein lacking most, or all, of the conserved ED (tyrosine phosphatase) domain. There is some evidence that while loss of just the tyrosine phosphatase domain causes non-syndromic sensorineural hearing loss, disruption of both the ED and the N-terminal domains is associated with syndromic (dilated cardiomyopathy with sensorineural hearing loss) disease [75,78]. One clinically actionable extension of these studies on the pleiotropic effects of Eya4 gene variants is described in a recent study where a ‘genotype-first’ approach was used to identify as-yet undiagnosed cardiac conduction disorders in adults presenting with hearing loss [80].

A recurrent mutation in Eya3 (resulting in an Asn358Ser substitution) causing oculo-auriculo-vertebral spectrum has recently been reported [81]. This mutation appeared to increase the half-life of EYA3 but have no impact on tyrosine phosphatase activity. This report notwithstanding, in general Eya2 and Eya3 mutations are rarely been linked with developmental disorders. This could be a consequence of redundancy, since while both genes are broadly expressed during development, they are most often co-expressed with other Eya genes.

Many decades after Drosophila EYA was first described, insights into the roles of EYAs in normal development, physiology and disease continue to emerge. Not described here, but of emerging clinical importance, is the therapeutic potential of targeting the EYA PTP activity for diseases ranging from cancer [82–84] to pulmonary arterial hypertension [85] and proliferative retinopathies [35]. Much remains to be clarified regarding EYA protein function and regulation. The question of whether the N-terminal domain and C-terminal domain activities of the EYA proteins impact on each other is still a matter of some debate. The fact that interaction with the SIX proteins is mediated by the ED domain, while transactivation potential exists in the N-terminal domain suggests that, at least for the transactivation activity in the context of the RDGN, the two domains are functionally linked. It is curious that the PTP activity of the EYA proteins has not yet been linked to any major developmental role, other than contributing to post-natal vascular development in the murine retina [35]. This is particularly notable since the EYA-PTP activity promotes DNA damage repair [10] and cell motility [14,15], two processes important in development: embryonic development involves rapid cell proliferation inevitably accompanied by DNA damage, and cell motility plays a crucial role in morphogenesis. It is likely that subtle developmental defects attributable to the PTP activity will be uncovered upon more careful examination in the future.

How did the EYA proteins come to have so many biochemical activities in a single polypeptide? It is interesting to speculate about where these activities originated. Notably, plants have an EYA protein that is comprised solely of the ED domain [7], and no significant amino-acid sequence homology exists between the EYA N-terminal domain and any other known protein sequence [4]. It has been suggested that some animal EYA proteins with an ED-like domain have very weak, or no, tyrosine phosphatase activity [86]. But this has not been experimentally queried. Could the EYA proteins we see in vertebrates have emerged from an ancestral ED domain-containing protein that may, or may not, have had tyrosine phosphatase activity? Does the N-terminal domain have intrinsic threonine phosphatase activity? Evolutionary biochemists suggest that all enzymes are promiscuous (have an average of 10 promiscuous activities [87–89]). Perhaps the putative threonine phosphatase activity of the EYA proteins is one such instance of evolution in progress? In this review, we have attempted to match the observed biochemical activities of the EYA proteins with their developmental roles, as understood from our vantage point at one moment in the evolutionary journey of the EYA proteins.

• EYA transactivator-phosphatase proteins play crucial roles in embryonic development and in tissue homeostasis. Mutations or mis-expression of EYA are linked to human genetic disorders and cancers, offering opportunities for the development of therapeutics and genetic diagnoses.

• The vast majority of the studies thus far implicate the EYA transactivation function in cell-fate determination. However, a growing body of evidence suggests that the tyrosine phosphatase activity (and the putative threonine phosphatase activity) could contribute to disease states. These observations are now being exploited in efforts to devise EYA-targeted therapeutics.

• Beyond the further examination of the myriad signaling pathways impacted by EYA-mediated transactivation and dephosphorylation events, examination of the molecular evolution of this unusual protein is of great interest. Such studies could be important contributors to our understanding of how phosphorylation-based signaling mechanisms evolved in general.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported through grants from the NIH NCI RO1-CA207068 and NHLBI HL152094 to RSH.

US helped write, edit, and generate figures, K.R. helped write and edit, R.S.H. wrote this article and generated figures.

BOR

branchio-oto-renal

EYA

Eyes Absent

NICD

Notch intracellular domain

OFC

otofaciocervical syndrome

PP2A

protein phosphatase 2A

PTP

protein tyrosine phosphatases

RDGN

retinal determination gene network

TSH

thryroid stimulating hormone