Comparative Functional Analysis of ZFP36 Genes during Xenopus Development

ZFP36 constitutes a small family of RNA binding proteins (formerly known as the TIS11 family) that target mRNA and promote their degradation. In mammals, ZFP36 proteins are encoded by four genes and, although they show similar activities in a cellular RNA destabilization assay, there is still a limited knowledge of their mRNA targets and it is not known whether or not they have redundant functions. In the present work, we have used the Xenopus embryo, a model system allowing gain- and loss-of-function studies, to investigate, whether individual ZFP36 proteins had distinct or redundant functions. We show that overexpression of individual amphibian zfp36 proteins leads to embryos having the same defects, with alteration in somites segmentation and pronephros formation. In these embryos, members of the Notch signalling pathway such as hairy2a or esr5 mRNA are down-regulated, suggesting common targets for the different proteins. We also show that mouse Zfp36 protein overexpression gives the same phenotype, indicating an evolutionary conserved property among ZFP36 vertebrate proteins. Morpholino oligonucleotide-induced loss-of-function leads to defects in pronephros formation, reduction in tubule size and duct coiling alterations for both zfp36 and zfp36l1, indicating no functional redundancy between these two genes. Given the conservation in gene structure and function between the amphibian and mammalian proteins and the conserved mechanisms for pronephros development, our study highlights a potential and hitherto unreported role of ZFP36 gene in kidney morphogenesis.


Introduction
Zinc-finger-containing-proteins constitute the most abundant protein superfamily in eukaryote genomes and they are involved in various cellular processes through their binding to DNA, RNA or protein [1]. Among this super family are subfamilies of proteins containing a variable number of zinc finger motifs based on a cysteine-histidine repeat with the configuration cys-cys-cys-his (C3H) [2]. One subclass of this family contains proteins that possess two C3H type zinc finger domains Cx 8 Cx 5 Cx 3 H (where x is a variable amino acid) or a Tandem Zinc Finger domain (TZF) separated by an 18 amino acids linker region. The prototype of this family is named ZFP36, previously described as TIS11, Tristetraprolin (TTP), Nup475 and GOS24 and which is rapidly induced by several mitogens [3,4,5,6,7]. Depending on the species, two or three other ZFP36 genes have been found in vertebrates. In human, in addition to ZFP36, there are two other genes namely ZFP36L1 and ZFP36L2. In rodents a fourth gene, Zfp36L3, has been identified and shown to be expressed only in placenta [8]. ZFP36 proteins have been showed to bind to AU rich elements (ARE) present in the 39UTR region of several mRNA encoding cytokines like Tumor Necrosis Factor alpha (TNFa) or the Granulocyte-Macrophage Colony-Stimulating Factor (GM-CSF) and this binding involves the tandem zinc finger domain of the proteins [9,10,11,12,13]. As a consequence, mice deficient for Zfp36 by gene targeting although appear normal at birth, soon develop a complex syndrome related to medullar and extramedullar myeloid hyperplasia associated with an increased cellular concentration of TNFa mRNA [14]. Inactivation of Zfp36l1 gene in mouse by knockout leads to the death of the embryo in utero at about 11 days of development by failure of chorioallantoic fusion, the embryos showing extraembryonic and intraembryonic vascular abnormalities along with heart defects [15,16]. Mutation of Zfp36l2 gene in the mouse causes female infertility and together, these knockout studies suggest distinct and non redundant functions for ZFP36 genes during development [17]. Mice lacking Zfp36l1 and Zfp36l2 genes during thymus development are prone to acute lymphoblastic leukemia and show elevated Notch1 mRNA levels [18], illustrating the importance of those RNA binding proteins during organ development and homeostasis.
Members of the ZFP36 gene family have been identified in several metazoans such as Drosophila, zebrafish and more recently in mollusc [19,20,21,22,23]. In the amphibian Xenopus laevis, four distinct genes that code proteins containing a tandem zinc finger domain have been identified and named xC3H-1 to 4 [24,25]. XC3H-1, xC3H-2 and xC3H-3 genes are true orthologs of the human ZFP36, ZFP36L1 and ZFP36L2 genes respectively. xC3H-4 is distinct from other ZFP36 genes, being unique to amphibians and encoding a protein with two tandem zinc fingers instead of one [24,25]. In agreement with Xenopus Gene Name guidelines, we will refer xC3H-4 to zfp36l4 and use zfp36, zfp36l1 and zfp36l2 for the other members of the family. Zfp36l1 and zfp36l2 have been showed, either by gain-of-function (for zfp36l1) or by gain and lossof-function (for zfp36l2) to be involved in Xenopus pronephros formation while zfp36l4 has been shown to regulate meiosis [25,26,27]. However, no functional study has been performed yet on zfp36, the prototype of the family and no comparative functional study between the different zfp36 proteins has been undertaken.
To gain more insight into the evolutionary history of ZFP36 genes, we have compared in detail their genomic structure between various metazoan phyla and found that vertebrates and basal metazoan ZFP36 genes are structurally conserved while protostome genes have diverged. In order to complete our knowledge on the amphibian zfp36 gene family, we have analyzed the developmental expression of zfp36 gene and performed a functional analysis. We found that the amphibian zfp36 gene has a unique expression pattern during development, one that is associated with somitic segmentation and nephrogenesis. When overexpressed in Xenopus embryos, each member of the zfp36 gene family gives the same embryonic defects suggesting common targets to all members of the family. We have identified several mRNAs whose expression is abolished or strongly reduced when the different zfp36 mRNA are overexpressed and in morphant embryos. Because zfp36 proteins are potential regulator of mRNA deadenylation and translation we may hypothesize they act on those mRNAs to regulate an early phase of organogenesis.

Results
The structural organization of ZFP36 genes is conserved between evolutionary distantly related animals Genes encoding proteins containing two C3H type zinc finger domains (Cx 8 Cx 5 Cx 3 H) (or TZF for Tandem Zinc Finger) have been independently cloned by several groups and identified by a variety of names (see introduction). In accordance with recommendations of the HUGO Gene Nomenclature Committee (http://www.genenames.org/), we propose to use ZFP36 as the founder name for members of this family, in place of the previous designations Tis11 or TTP. Therefore, in addition to the three human genes ZFP36, ZFP36L1 and ZFP36L2, the fourth gene identified in rodents and belonging to this family is named Zfp36l3. Among the four genes containing a TZF domain described in Xenopus, xC3H-1, xC3H-2 and xC3H-3 are true orthologs of the human ZFP36, ZFP36L1 and ZFP36L2 as confirmed by synteny and phylogenetic analyses ( Fig. 1A and Fig. S1). The fourth Xenopus zfp36 gene, xC3H-4, is distinct from other ZFP36 genes and unique to amphibian genome [24,25]. In agreement with Xenopus genes names guidelines, we will refer xC3H-4 to zfp36l4 and use zfp36, zfp36l1 and zfp36l2 for the other members of the family.
ZFP36 genes have been identified in numerous metazoans but no comparative analysis based on gene structure has been performed yet on those genes. In order to obtain a more comprehensive picture of the ZFP36 gene family in metazoan, we searched in EST and genomic databases for the presence of genes containing a conserved TZF domain in several metazoan taxa. We identified a unique gene containing a TZF domain in the sea anemone Nematostella vectensis that belongs to the basal metazoan cnidarians. A single gene coding a TZF containing domain was also identified in the genome of the sponge Amphimedon queenslandica, a member of an ancient group of animals that has diverged from other animals over 600 Ma. Similarly we found a single ZFP36 gene in two basal deuterostomes bilateria, the echinoderm sea urchin Strongylocentrotus purpuratus and the urochordate Ciona intestinalis. A phylogenetic tree made with the TZF domain encoded by these different genes indicates that ZFP36, ZFP36L1 and ZFP36L2 have evolved from a single gene and that ZFP36L1 and ZFP36L2 are closely related and probably resulted from a duplication event during evolution (Fig. S1).
In order to gain insights into the evolution of the zfp36 genes, we retrieved their genomic organization for different taxa and compared with with the vertebrate ZFP36, ZFP36L1 and ZFP36L2 genes. The human, mouse and amphibian genes showed a highly conserved organization with two exons separated by a phase 0 intron of variable size (Fig. 1B). When compared to the situation in other taxa, one striking finding is that the gene structure observed in vertebrates extends not only to other deuterostomes (sea urchin and ciona) but also to basal metazoans such as cnidarians (Nematostella) and sponges (Amphimedon). In each case, the two exons are separated by a phase 0 intron and the TZF domain is always found in the second exon (Fig. 1B).
Zfp36l3 and zfp36l4 have a different gene structure from other Zfp36 genes (data not shown). In rodents, Zfp36l3 is intronless, suggesting that the gene has arisen in that lineage by retrotransposition of a processed cDNA (data not shown). In Xenopus, zfp36l4 gene is constituted by two exons, but exon 1 contains only 59 untranslated sequence (data not shown). This feature is also compatible with a retrotransposition event that may have occurred in the amphibian lineage.
In contrast to other taxa analysed, zfp36 gene structure differs in protostomes. The unique Drosophila zfp36 gene contains three exons separated by a phase 0 intron and a phase 2 intron respectively (data not shown) and the nematode Caenorhabditis elegans zfp36 (ccch-1) comprises 9 exons (data not shown). Since the unique intron in the vertebrate ZFP36, ZFP36L1 and ZFP36L2 genes is found at the same position and also in the basal orthologous genes of cnidarians and sponges, we conclude that this splice structure is an ancestral trait. The more complex gene structure observed in Drosophila and C. elegans zfp36 genes reflects secondary lineage-specific gain of introns. Together this analysis reveals a strong conservation in the structure of the ZFP36 gene in the deuterostome lineage. The three vertebrates genes ZFP36, ZFP36L1 and ZFP36L2 share the same structural organization and this is compatible with the gene duplication events that have occurred during vertebrates evolution [28].

Comparative analysis of zfp36 genes expression pattern during Xenopus development
The cloning and expression of members of the Tis11/TTP gene family in Xenopus has been previously reported [24,25,27], however these studies were primarily focused on zfp36l1 or zfp36l2 but not on zfp36 (Tis11/TTP) the founding member of the family. Moreover, no gene function study has been yet performed on zfp36. Before undertaking functional studies, we further evaluated zfp36 expression during development and performed a detailed expression of zfp36 as compared to other gene family members. RT-PCR analysis indicates that zfp36l, zfp36l1 and zfp36l2 genes are expressed maternally and throughout development at a constant level from egg to tadpole stage ( Fig. 2A). In contrast, there is a decrease of zfp36l4 mRNA level after fertilization, with complete disappearance after midblastula transition ( Fig. 2A).
These data confirm previous observations obtained by Northern blot for the four genes and by RT-PCR for zfp36l2 [24,25]. Because the four zfp36 genes are maternally expressed, we analyzed the localization of their corresponding mRNAs. In situ hybridization combined with histological sections revealed that the four genes are expressed at the animal pole in 4-cell stage embryo   Fig. 2B, a-h). In the morula embryo, zfp36 mRNA is concentrated at the animal pole while the three other mRNAs spread from the animal pole to the marginal zone (Fig. 2B, i-l). mRNA distribution was then analyzed at blastula stage by RT-PCR on dissected embryos. When the blastula embryo is dissected into animal versus vegetal pole, zfp36, zfp36l1 and zfp36l4 mRNAs are mostly found in the animal pole region while zfp36l2 mRNA is also expressed in the vegetal pole (Fig. 2C). At the gastrula stage, after zygotic transcription had resumed, zfp36, zfp36l1 and zfp36l2 mRNAs are found in the animal pole, but also in ventral and dorsal mesoderm, indicating ubiquitous expression (Fig. 2D). As expected, zfp36l4 is not expressed in gastrula stage embryo (Fig. 2D).
The spatial expression of the four zfp36 genes was then analyzed on later stages embryo by whole mount in situ hybridization. Zfp36 mRNA is detected in the somites, cement gland and appears as punctate staining in the lateral mesoderm of the stage 24 embryo (Fig. 3a). By stage 28, zfp36 expression is restricted to the cement gland and appears in the notochord where it extends to its rostral end (Fig. 3b). In the stage 33/34 embryo, expression persists in the cement gland but notochord expression is no longer detected while the pronephric tubule is stained as seen in a close view section (Fig. 3c). Zfp36l1 is mainly expressed in pronephros anlagen and brain of stage 24 embryos and at a lower level in somites (Fig. 3d). By stage 28 embryo, pronephros expression persists and additional sites of expression are detected in the midbrain, otic vesicle and branchial arches (Fig. 3e). At stage 35/36, the pronephros tubule and duct are both stained (Fig. 3f) as previously reported [27]. Zfp36l2 expression is detected in pronephros anlagen in the stage 24 embryo where it persists through stage 28, with additional expression in branchial arches, otic vesicle and midbrain (Fig. 3g, h). By stage 35/36, Zfp36l2 expression is detected in the pronephric duct and surrounds the pronephric tubule ( Fig. 3i) as previously reported [25]. Zfp36l4 mRNA is not detected in stage 24 nor at later stages ( Fig. 3j-l) thus confirming our RT-PCR data ( Fig. 2A) and previous Northern blot analysis [24].
Taken together, these results indicate that zfp36 has a unique expression pattern when compared to other zfp36 gene family members. Although zfp36 is expressed during pronephros development like zfp36l1 and zfp36l2, it is a late marker compared to the two other genes and its expression is restricted to the pronephric tubule whilst zfp36l1 and zfp36l2 also mark the pronephric duct.

Different effects of signalling pathways on zfp36 expression in embryonic cells
The mammalian ortholog of Xenopus zfp36 is a primary response gene that is rapidly and transiently induced in fibroblasts when treated with serum and several mitogen factors such as FGF, PDGF or insulin [3,5]. To determine whether the expression of zfp36 gene family members was modulated by growth factors, we used the animal cap assay and tested the effects of FGF, activin and BMP that are major signalling pathways acting in the early embryo [29]. FGF2 treatment of animal caps resulted in a dosedependent increase in expression of zfp36, zfp36l1 and zfp36l2 (Fig. 4A). This increase is specific and is not observed in the presence of SU5402 (SU), an inhibitor of the FGF signalling pathway (Fig. 4A). Activin treatment stimulated the expression of zfp36l1 and zfp36l2 but had no effect on zfp36 expression (Fig. 4B). The effect of activin on zfp36l1 and zfp36l2 expression is not observed in the presence of SB431542 (SB), an inhibitor of activin signalling pathway (Fig. 4B). Animal caps derived from BMP2 mRNA-injected embryos showed a decrease in zfp36l1 expression and an increase in zfp36l2 expression, but no difference in zfp36 expression whatever the amount of injected mRNA (Fig. 4C). In all those experiments, zfp36l4 was never expressed and proved totally unresponsive to growth factor treatments. These data indicate that the different zfp36 genes respond differently to the major signalling pathways that are active in the early embryo.

Gain-of-function of zfp36 genes causes somites defects
As a first step towards understanding the functions of zfp36 in the early embryo, we expressed the mouse or the amphibian proteins through the injection of their corresponding mRNA. In preliminary experiments, different amounts (from 50 pg to 1 ng) of zfp36 mRNA were injected into one blastomere of two-cell stage embryos. Injected embryos developed a typical curved axis phenotype, with severity increasing at higher dose. Over 80% (n = 115) of the embryos injected with 250 pg exhibited this highly penetrant phenotype, suggesting a trunk elongation alteration that could be related to alteration in somites (data not shown). To further evaluate somite development, injected embryos were analyzed at tailbud stage by immunhistochemistry with the somite specific marker 12/101. Embryos injected with mouse or amphibian zfp36 mRNAs showed a slight decrease in 12/101 staining on the injected side compared to the control uninjected side in about 80% of embryos (n = 55) ( Fig. 5a, b, d, e). Moreover, the injected side did not show the typical blocks of regularly spaced somites but instead a uniform labeling (Fig. 5b, e). This is more obvious in histological sections (Fig. 5c, f). The ultrastructural defects of somites were confirmed by electron microsocopy analysis (Fig. 5g, h). The injection of zfp36l1, zfp36l2 or zfp36l4 mRNAs resulted in the same phenotype ( Fig. 5i-n). In these experiments, notochord was not apparently altered ( Fig. 5c or 5f) a finding confirmed by immunohistochemistry with the specific antibody MZ15 [30] (data not shown).
The reduction in somite staining with 12/101 antibody observed in some cases could result from an impairment in mesoderm induction and/or myogenic differentiation. In order to establish whether zfp36 mRNA overexpression could affect mesoderm induction, animal cap explants derived from microinjected embryos were treated with activin and assayed for the expression of the pan-mesodermal marker brachyury (xbra). Animal cap cells derived from microinjected embryos showed no notable change in the expression of xbra, indicating that mesoderm induction is not affected by zfp36 gene expression levels (Fig. 6A).
We observed identical results when zfp36l1, zfp36l2, zfp36l4 or mouse zfp36 mRNAs are injected (Fig. 6A). To evaluate the effect of zfp36 overexpression on myogenic differentiation, injected embryos were analysed by in situ hybridization for the expression of the myogenic regulatory gene myod. Embryos injected with Xenopus zfp36 mRNA showed a more diffuse myod expression on the injected side with an altered segmentation pattern compared to the uninjected side (Fig. 6B, a, b). The same results were observed when zfp36l1, zfp36l2, zfp36l4 or mouse zfp36 mRNAs were injected (Fig. 6B, c-f and data not shown). Together these results indicate zfp36 gain-of-function does not affect mesoderm induction or myogenic differentiation but rather impairs somite segmentation.

Alteration of esr5 and hairy2a expression pattern
The Notch signaling pathway is central to somitogenesis by controlling somite segmentation through downstream components like esr5 and hairy2a [31,32]. In order to know whether members of the Notch pathway were affected in gain-of-function experiments, we analysed the expression of esr5 and hairy2a in embryos injected with zfp36 mRNA. Esr5 is expressed in a posterior tailbud domain of the embryo marking the presomitic mesoderm, and also in the anterior part of the first two somitomeres, forming two chevrons (Fig. 7a) [32]. Unilateral injection of zfp36 mRNA resulted in embryos showing no clear demarcation between the two chevrons and in some cases only one chevron was detected (72% n = 45) (Fig. 7b). The same phenotype was observed with embryos injected with zfp36l1, zfp36l2 or mouse zfp36 mRNAs (Fig. 7c-h). In the tailbud embryo, hairy2a expression is normally found in the presomitic mesoderm as a chevron-shaped stripe and also in the pronephros anlagen (Fig. 7i, k). Embryos injected with either Xenopus or mouse zfp36 mRNA showed a reduced expression of hairy2a in both the somites and pronephros anlagen (78% n = 50) (Fig. 7j, l). We conclude from these data that overexpression of zfp36 can indeed affect the Notch signaling pathway.

Pronephros defects
Previous studies have shown that overexpression of either zfp36l1 or zfp36l2 mRNA in Xenopus embyo induces pronephros, abnormalities mainly affecting morphological development of pronephric tubules [25,27]. Because zfp36, like zfp36l1 and zfp36l2, is expressed in pronephros and can alter hairy2a expression in the pronephros anlagen (see Fig. 7j), we analysed whether its overexpression could affect pronephros development. Embryos injected at the 8-cell stage in ventral balstomere with zfp36 mRNA were allowed to develop until tadpole stage and then whole-mount immunostained with the monoclonal antibodies 3G8 and 4A6, markers of pronephric proximal tubule and pronephric duct respectively [33]. Unilateral injection of zfp36 mRNA resulted in a reduction of tubule and duct staining in embryos (85% n = 35) (Fig. 8A, a-d). Microinjection of zfp36l1 or zfp36l2 mRNA gave reproducibly an alteration of pronephros development, thus confirming published data (Data not shown) [25,27]. In these experiments, we consistently found that the somites were not altered when analysed by immunohistochemistry with the somite specific marker 12/101 antibody, confirming a direct effect of zfp36 overexpresssion on pronephros development (data no shown).
To further investigate the effect of zfp36 overexpression on pronephros development, we analyzed the expression of pax8 and lim1, key players in pronephros development [34,35]. Injection of zfp36 or zfp36l1 mRNAs resulted in a marked reduced expression of pax8 and lim1 (80%, n = 55) (Fig. 8B, a-h). Embryos injected with zfp36l2 mRNA showed a reduction of pax8 expression but no change in lim1 expression (data not shown) as previously reported [25].
To confirm a role of zfp36 in pronephros development, we used the pluripotent animal cap cells model where pronephros tissue can be induced by treatment with activin and retinoic acid [35,36,37]. Animal caps derived from embryos injected with the different zfp36 mRNAs were assayed for the expression of SMP30, a specific marker of pronephric tubule that is induced under conditions where pronephros is formed [38]. A strong inhibition of SMP30 expression is observed in animal caps injected with zfp36 mRNAs when compared to non injected control caps (Fig. 8C). We conclude that overexpression of zfp36 affects pronephros development.

Loss-of-function of zfp36 affects pronephros morphogenesis
Zfp36l1 and zfp36l2 have been previously shown to affect kidney development either in gain-of-function (for zfp36l1 and zfp36l2) or loss-of-function experiments (for zfp36l2) [25,27]. Zfp36 which is also expressed in pronephros (See Fig. 3c) may be potentially involved in its development. We tested this hypothesis by loss-offunction analysis using a morpholino antisense (MO) knockdown assay. We also tested similarly the effects of zfp36l1 knockdown because only gain-of-function has so far been described for this gene [27]. We designed two morpholinos to interfere specifically with the translation of each mRNA. The efficiency of MOs to inhibit translation of their respective mRNAs was established in an in vitro reticulocyte lysate system and in vivo in the embryo (Fig. S2 and data not shown). MOs were injected in 8-cell stage embryos to target pronephros anlagen and the development of the pronephros was evaluated in tadpole embryos by immunohistochemistry with 3G8 and 4A6 antibodies. Knockdown of either zfp36 and zfp36l1 resulted in a similar phenotype characterized by alteration of pronephric tubule morphology, while injection of control MO had no effect (Fig. 9a-f). The phenotype of zfp36 morphants ranged from a mild to strong size reduction of the tubule, with a defective coiling of the duct (84%, n = 56) (Fig. 9b). Zfp36l1 morphants also showed alteration of tubule size but, unlike zfp36 morphants, they also displayed alteration of the distal portion of the duct which is missing (80%, n = 66) (Fig. 9d). Magnified views of the injected side of selected morphant embryos clearly show alteration of tubule morphogenesis (Fig. 9i-p). Morphant embryos also developed edemas, visible from stage 42, a possible consequence of impaired pronephric function [39] (data not shown). The pronephros defects in zfp36 morphants can be rescued by coinjection of 100 to 200 pg of mouse zfp36 mRNA (70%, n = 36) indicating that the knockdown effect was specific (Fig. 9g, h).
It is known that signals from anterior somites are involved in pronephros formation. Therefore, we assessed whether MO depletion of zfp36 could affect the formation of paraxial mesoderm formation, thus contributing indirectly to pronephros defects. 4cell stage embryos were injected dorsally with zfp36 MO and then analysed at stage 15 for myod expression by in situ hybridization and at stage 30 by immunohistochemistry with an antibody (12/ 101) that specifically labels the somites. Myod expression was unchanged in morphants embryos (n = 18) and the labeling of somites with 12/101 antibody was not affected, showing the typical regularly spaced chevron-like structures (n = 23) (Fig. 10A). Taken together, we conclude that targeting zfp36 morpholinos to mesoderm does not alter somite formation and thus the depletion of zfp36 is likely to have a direct effect on pronephros development.
In order to characterize in more detail the phenotype of zfp36 morphants, we next asked whether pronephros specification was affected by loss of zfp36 expression. Several genes have been shown to be implicated in early specification and development of the pronephros, including lim1, pax8, wnt4 or wt1 [34,40,41,42,43]. We did not observe any change in the expression of lim1 (n = 31)e, pax8 (n = 33), wnt4 (n = 25) or wt1 (n = 22) in zfp36 morphants (Fig. 10B) indicating that zfp36 is not required for early specification of kidney.
Notch signaling has been shown to play an important role during pronephros development [40,44,45]. As components of this pathway are potential zfp36 targets, we have tested whether hairy2a expression, a downstream effector of the pathway, was affected in morphant embryos. Embryos injected with zfp36 MO were cultured until stage 33/34 and analysed for hairy2a expression. We did not observe any change in the expression of hairy2a (n = 25) indicating that Notch signaling pathway was not affected by zfp36 depletion.
The pronephric defects in zfp36 morphants could be explained by a decrease of proliferation or the apoptotic elimination of pronephric cells. To evaluate these possibilities, embryos were injected at the 8-cell stage in prospective pronephric territory with zfp36 MO and then analysed for proliferation by immunohistochemistry with an antibody recognizing phosphorylated-histone H3, and for apoptosis by TUNEL. No significant alteration of proliferation (n = 25) or apoptosis (n = 32) was observed on the injected side compared with the uninjected side (Fig. S3). We conclude that the pronephric defects observed in zfp36 morphants are not related to changes in proliferation or apoptosis.

Discussion
The present work extends our knowledge on zfp36 genes and provides a functional comparison between the vertebrate members of this family using Xenopus as an experimental model. Zfp36 genes have evolved from a single gene present in basal metazoans. Indeed, in our survey of extant animals genomes containing a Tandem Zinc Finger domain identical to human ZFP36, we have identified a single gene present in the genome of basal phylum such as porifera (Amphimedon) and cnidaria (Nematostella). There is also a single zfp36 gene in protosotomes (Drosophila, C.elegans or mollusc) and in basal deusterostomes such as echinoderm (sea urchin) or urochordates (ciona). In contrast, there are three to four zfp36 genes in vertebrates, depending on the species, in agreement with the hypothesis that two rounds of duplication have occurred between the origin of chordates and the origin of jawed vertebrates [28]. Of the four zfp36 genes present in rodent and amphibian, three of them, ie zfp36, zfp36l1 and zfp36l2, are true orthologs while the fourth one, zfp36l3 in rodent and zfp36l4 in amphibian has probably arisen through retrotransposition of a processed cDNA. When considering gene structure, one striking finding is that there is strong conservation between the unique zfp36 gene found in sponges and the deuterostome zfp36 orthologs, where the unique intron is found at the same position and in the same phase. Moreover, in all genes analyzed, the TZF domain is constantly found in the second exon. This conservation, that has been maintained over more than 800 My of evolution, is not found in protostomes genomes of Drosophila melanogaster and C. elegans where zfp36 genes show an increased number of introns. This suggests that the common ancestor of bilaterian zfp36 had a simple gene structure with two exons and that the variation in gene number (observed for example in Drosophila melanogaster or C.elegans) reflects secondary lineage specific gain of introns [46].
We have used the Xenopus embryo to evaluate whether different zfp36 genes could have distinct functions during early development. Our work demonstrates that FGF, activin and BMP, some of the major signaling pathways during early Xenopus development, are implicated at some level in the differential regulation of zfp36 in the embryo. Zfp36 are RNA binding proteins that recognize AU-rich element within 39UTR of mRNAs leading to their deadenylation and accelerated degradation [7]. We may therefore hypothesize that their function is closely related to their temporal and spatial expression. zfp36l4 is maternally expressed and not detected after midblastula transition and therefore has probably no function in early development and organogenesis.  (e, f) or mouse zfp36 (mzfp36, g, h, k, l) mRNAs were injected into one blastomere of two-cell stage embryos and developing embryos were fixed at stage 25 (a-h and k, l) or stage 28 (i, j) and analyzed by in situ hybridization for esr5 and hairy2a expression. Arrows in a, c, e and g mark the presomitic mesoderm (psm) and the pronephros region (pn) in i-l respectively. Stars in a, c, e and g mark the first two somitomeres. doi:10.1371/journal.pone.0054550.g007 Indeed, it has been shown to be required for meiosis progression [26]. Zfp36l1 and zfp36l2 expression patterns are closely similar in the embryo [25,27]. Although zfp36 is expressed in pronephros like zfp36l1 and zfp36l2, it differs from the two other, being the sole member of the family expressed in the notochord. Moreover, when considering pronephros development, zfp36 expression is delayed when compared to the two other genes. This expression pattern could reflect distinct regulatory elements that are shared by zfp36l1 and zfp36l2 and their common origin through genome When overexpressed in the embryo, each of the four zfp36 proteins induces somite segmentation defects and pronephros alteration without affecting myogenesis or somite formation. In agreement with the indispensable requirement of Notch signalling for somite segmentation, we have found that the expression of esr5 and hairy2a mRNAs, two members of the signalling pathway, is altered. Those mRNAs possess an ARE in their 39UTR and therefore can be targeted for degradation by zfp36 proteins (data not shown). A survey of mRNA possessing ARE elements in their 39UTR indicates that several members of the Notch signaling pathway, such as Notch itself, Delta or esr9 mRNAs, could be direct targets of zfp36 proteins (data not shown). The direct involvement of ZFP36 in Notch signalling has been recently described in  human thymocyte development where a direct interaction between ZFP36L1 or ZFP36L2 and an ARE present in the 39UTR of Notch1 mRNA has been observed [18]. A striking finding is that the three predicted binding sites for ZFP36 in the mammalian 39UTR Notch1 mRNA are totally conserved in the Xenopus 39UTR notch mRNA (data not shown). Among the three binding sites, there is a nonamer sequence UUAUUUAUU that has been described as the optimum binding site for all zfp36 family members [11,47,48]. Interestingly, this sequence is also found in the Xenopus hairy2a 39UTR reinforcing the idea that several members of Notch pathway signalling can be targeted by zfp36 proteins. Surprisingly, this nonamer sequence is also found in the 39UTR region in WT1 (Wilms tumor suppressor) mRNA that encodes a protein that has been shown to be involved in nephron defects in vertebrates [49]. However, we cannot exclude the possibility that alteration of somite segmentation may also be indirect and a consequence of the targeting by zfp36 of mRNAs that themselves my regulate somite segmentation. For instance, one such target could be EDEN BP mRNA whose downregulation disrupts esr5 expression pattern in a way similar to zfp36 [50]. Several other mRNA encoding RNA binding proteins have also been shown to regulate somitogenesis and can be potentially targeted by zfp36 [51,52].
Notch activation is essential for pronephros development both in Xenopus and in mouse, acting on proximal tubule and glomus formation [40,44,53,54,55]. We can hypothesize, from our gainof-function experiments, that zfp36 acts through the targeting of Notch signalling pathway elements, like hairy2a or esr5, to alter pronephros formation. Zfp36 can also act at an early step in pronephros development and target important regulators such as pax8 or lim1 mRNA whose expression is significantly decreased in zfp36 overexpressing embryos. These data confirm previous observations made on zfp36l1 and zfp36l2 [25,27]. During this work, we have constantly found that the mouse Zfp36 protein induces the same embryonic defects than the amphibian proteins, suggesting a functional conservation between vertebrate proteins. We have a limited knowledge of mRNA targets of zfp36 family members and whether individual members of the family target distinct, overlapping, or identical targets to other family members. Our gain-of-function experiments provide evidence that the distinct zfp36 proteins give the same somite segmentation and pronephros defects, suggesting that the different zfp36 proteins may have the same mRNAs targets.
Loss-of-function using morpholino antisense oligonucleotides interfering with zfp36 translation shows an extremely consistent phenotype characterized by a dramatic alteration of pronephros development leading to tubule size reduction and coiling defects. Those defects, that are not related to somites alteration can be rescued by the mouse zfp36 protein, indicating that they are specific but also that there is a strong functional conservation between the amphibian and the mammalian proteins. The expression of genes involved at early stages of pronephros formation such like lim1, pax8, wnt4 or wt1 is not affected in zfp36 morphants embryos, nor is apoptosis or proliferation. Notch pathway, at least its downstream effector hairy2a, seems not also affected by zfp36 depletion. We conclude from our experiments that zfp36 expression is not essential at early step of the pronephros specification, but is critically required at a later step of its organogenesis. We have shown for the first time that this is also the case for zfp36l1. Our data extend previous work that showed that zfp36l2 knock down impaired pronephros development [25]. Therefore, from our report and from published work, all zfp36 gene family members that are zygotically expressed in the early Xenopus embryo are necessary for a correct pronephros development. One striking issue that emerges from morphant analysis is the complete lack of redundancy between those genes while overexpression of all members induce the same phenotype. In over expression studies, zfp36 proteins are expressed in all regions of the embryo and therefore they can target the same mRNA as long as they possess ARE in their 3-UTR. We have shown this can be the case for several Notch signaling members. In the case of knockdown, the lack of redundancy between zfp36 family members could be explained by the distinct spatial and temporal expression of each gene. Zfp36l1 and zfp36l2 are expressed in pronephros anlagen but not in a totally overlapping pattern, while zfp36 expression, in contrast to the two others, is temporally delayed and restricted to tubule expression. This suggests that, although they can target the same mRNAs, the different zfp36 proteins are not acting at the same development stage or in the same cells. However we cannot exclude that, even when co expressed in the same cells, the individual zfp36 proteins interact with distinct partners that are necessary for their function. One puzzling observation from our work is that both gain-offunction and loss-of-function strategies affect pronephros development in a similar way. We suggest that a fine balance between the different zfp36 proteins level is required for normal development. Nevertheless, gain-of-function and loss-of-function studies differ in the way they affect early steps in pronephros development. Whereas lim1 and pax8 expression is decreased in embryos overexpressing zfp36 proteins, their expression, like those of wnt4, or wt1 (two others major actors in pronephros development) is unchanged in zfp36 morphants embryos. We hypothesize there exist additional targets to be discovered, whose expression is finely tuned by zfp36 proteins at late stages of pronephros morphogenesis.
Although zfp36 proteins are predominantly located in the cytoplasm, an unexpected partner that has been described in the case of zfp36l1 is the transcription factor HNF1b which, when mutated, is responsible for kidney congenital defects [27]. In a search for mutations in the open reading frame of human ZFP36L1 in patients with renal anomalies none were found [27]. Because Zfp36l1 knock down in mouse is embryonic lethal, we may hypothesize that mutations in the human protein might also be deleterious for development. Since our results indicate that zfp36 and zfp36l1 have similar effects on pronephros development, it is conceivable that some human renal anomalies might be related to zfp36 mutations, thus opening new interesting investigations. Together our studies indicate that zfp36 gene family members have unique function during pronephros development and suggest a model in which they regulate late phase of organogenesis. While zfp36 proteins have previously been reported to be involved in inflammatory disease and cancer, our study establishes an additional critical role during kidney development and morphogenesis. Given the conservation in gene structure and function between the amphibian and mammalian proteins and the conserved mechanisms for pronephros development our studies have uncovered a potential role of zfp36 gene in human kidney disease that merits further investigation.

Ethics Statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the European Community. The protocol was approved by the ''Comité d'éthique en expérimentation de Bordeaux» Nu 33011005-A.

Plasmid constructions
The coding sequences of the four Xenopus laevis zfp36 genes have been cloned into pCS2 vector by PCR using primers containing restriction sites. Zfp36 cloning was made from the IMAGE clone 7009009 (Accession BC082435) while zfp36l1, zfp36l2 and zfp36l4 sequences were cloned by RT-PCR from embryo RNA. A flag epitope has been added to the C-terminal end of the sequences. Mouse zfp36 (mZfp36) coding sequence has been cloned by RT-PCR from Embryonic Stem cells RNA in pXT7 vector. The primers for cloning are as follows (cloning sites are underlined):

RT-PCR analysis
RT-PCR analysis was performed as previously described [56] with the following primer pairs: For an accurate semi quantitative analysis, we have used conditions where the signals obtained are a linear function of the input cDNA as measured by amplification of serial dilutions of the input cDNA (data not shown).

Microinjection and animal cap assay
Xenopus laevis eggs were obtained by injecting adult females with 750U human chorionic gonadotrophin. Staging of embryos was according to Nieuwkoop and Faber tables [57]. For induction assay, animal cap explants were dissected from stage 8.5-9 embryos and treated with different amount of bFGF, activin (R&D Systems) or a mixture of activin plus retinoic acid (SIGMA R2625) and cultured until the control embryos reached the appropriate stage before RT-PCR analysis. The FGF inhibitor SU5402 (SIGMA) and the activin inhibitor SB431542 (SIGMA) were used at 50mM. For microinjection experiments, we used a Nanoject system (Drummond Scientific) and the capped mRNAs were synthesized in vitro using Ambion mMessenger mMachine SP6 kit (Austin, TX). We determined in preliminary experiments the effective doses for the microinjection experiments corresponding to 250 pg zfp36 mRNA and 20 ng to 50 ng of MO. For rescue experiments, 100-200 pg of mouse Zfp36 mRNA were co-injected with the MOs. 250 pg of b-galactosidase mRNA were used as tracer and the injection were performed into one blastomere at either 2cell stage, 4-cell stage or in a ventral blastomere of 8-cell stage embryos. For animal cap assay, embryos were injected in the animal pole of 2-cell stage embryo into both blastomeres. Animal caps were then dissected at stage 8.5-9 and cultured to the appropriate stage before RNA extraction and RT-PCR analysis. All results shown are representative of at least two independents experiments.

Antisense morpholino oligonucleotide
Antisense morpholino oligonucleotides were obtained from Gene Tools LLC. The sequence of the antisense MO was based on the designed parameters recommended by Gene Tools, as follows: MOzfp36 (