The endosperm-specific ZHOUPI gene of Arabidopsis thaliana regulates endosperm breakdown and embryonic epidermal development

In angiosperm seeds, developing embryos are surrounded by the second zygotic product of double fertilization, the endosperm(Olsen, 2004). In many dicotyledonous plants, such as Arabidopsis, the endosperm is largely transient and is consumed during seed development. In Arabidopsis, a single specialized cell layer, which has been compared with the aleurone layer of monocotyledonous plants, is all that remains of the endosperm by seed maturity. Although the importance of the endosperm in embryo/seedling nutrition is uncontested, it may also provide developmental signals to the young embryo (Berger, 2003; Berger et al., 2006). In both monocotyledonous and dicotyledonous species, the region of the endosperm that directly surrounds the developing embryo (the embryo surrounding region or ESR) expresses a unique complement of genes, supporting the idea that it has a specialized function. Although several of these ESR-specific genes encode proteins with obvious metabolic roles, others encode small secreted peptides,some of which are related to the precursor of CLAVATA3 (CLV3), the likely ligand of the CLAVATA1 receptor-like kinase, supporting the supposition that the ESR may be involved in signalling(Opsahl-Ferstad et al., 1997; Sharma et al., 2003). However,no precise developmental role for any of these peptides has yet been demonstrated. Recent studies have also shown that the developing embryo can send developmental signals to the endosperm, as fertilization of the egg cell alone is sufficient to trigger proliferation of the central cell(Nowack et al., 2006).

One potentially important role that the endosperm could play in embryogenesis is in defining the continuous cuticular layer at the boundary between the developing embryo and the endosperm. This structure is necessary to prevent organ fusion after germination, and it may also prevent fusion of the embryo surface with surrounding endosperm tissues during seed development,thus allowing normal embryo growth. In addition, the presence of a watertight cuticle surrounding the mature embryo is crucial in protecting it from desiccation upon germination (reviewed by Jeffree, 2006). Although the production of cuticle is generally considered to be a strictly epidermal property, a cuticularized cell wall has been reported to surround the unicellular zygote in Citrus, well before a defined protodermal cell layer is formed (Bruck and Walker,1985). Interestingly, the expression of epidermal markers, such as the transcription factors ATML1 and PDF2, which are required for epidermal specification in Arabidopsis, is also detected well before protoderm formation (Abe et al.,2003; Lu et al.,1996). Thus, the outer layer of embryonic cells display epidermal characteristics, including the presence of cuticularized tissue, from an extremely early stage in embryogenesis.

Several recent publications have supported a role for the endosperm in embryonic cuticle production and the specification of embryonic epidermal identity in Arabidopsis. A secreted subtilisin-like serine protease ABNORMAL LEAF SHAPE1 (ALE1), which is expressed predominantly in the ESR appears to be required for normal cuticle production in Arabidopsisembryos (Tanaka et al., 2001). Plants that lack ALE1 produce embryos that adhere to the endosperm during development. Moreover, germinating seedlings are extremely sensitive to desiccation owing to abnormal cuticle production on their cotyledons. However,if ale1 seedlings are grown in vitro in high humidity, and then transferred to soil, they survive and their adult leaves have a normal cuticle structure, consistent with their cuticle defects being derived from abnormalities arising during embryogenesis. Mutations in either of two embryonically expressed genes encoding RLKs, ACR4 or ALE2,leads to a significant enhancement of the relatively mild ale1 phenotype. In double mutants, epidermal specification (as judged by the expression of the epidermal markers such as ATML1) is also partially lost(Tanaka et al., 2007; Watanabe et al., 2004). Neither acr4 nor ale2 mutants show a marked seedling phenotype, although both have cotyledon cuticle abnormalities, and the synergistic interactions of these genes may indicate that ACR4 and ALE2 are required to perceive a signal processed by ALE1. Such a function for ALE1 would also be consistent with the known role of several other subtilisin proteases in processing ligand signals(Berger and Altmann, 2000; Cui et al., 1998; Julius et al., 1984; Rose et al., 1996). Two other embryonically expressed receptor-like kinase encoding genes, GASSHO1and GASSHO2, also act redundantly to allow normal embryonic cuticle formation (Tsuwamoto et al.,2008), although how these genes interact genetically with ALE1, ACR4 and ALE2 has yet to be determined.

In this paper, we show that the transcription factor encoded by the ZHOUPI (ZOU) (Chinese for `shrivelled') gene of Arabidopsis is a key player in controlling ESR-expressed genes involved in the development of the embryo surface, including ALE1. The ZOU gene is widely conserved and occurs in plant groups that lack endosperm (gymnosperms) or seeds (Lycophyta). We propose that ZOUplays a pivotal role in the crosstalk between the embryo and surrounding nutritive tissues. In angiosperms, this is necessary to promote both cuticle formation and the separation of embryo from the surrounding tissues, whereas in more primitive plants it may permit invasive growth of embryos into nutritive tissue of the female gametophyte.

The semi-dominant zou-1D allele arose in a conditional curly leaf (clf) mutant background (clf-50 CLF::CLF-GR, Ws ecotype) described previously (Schubert et al., 2006). The mutant was backcrossed several times to Ws ecotype to transfer the zou-1D mutation into a CLF+ background and all experiments were performed using this line.

The recessive, loss of function zou2-4 alleles were obtained from the Nottingham Arabidopsis Stock Center. The zou2 allele (Ws background) arose in the FLAG collection of T-DNA insertions(Samson et al., 2002) and corresponded to FLAG 400A08. The zou-3 allele (Ws background) was obtained from the University of Wisconsin collection of T DNA insertion lines(Sussman et al., 2000) and corresponded to WiscDsLox465F5. The zou-4 (Col-0 background) allele was obtained from the Gabi-Kat collection of T-DNA inserts(Rosso et al., 2003) and corresponded to GABI_584D09. The position of the T-DNA inserts was confirmed by PCR amplification and sequencing of genomic DNA flanking the inserts. The ale2-1 and ale1-1 mutations were provided by Hirokazu Tanaka(Tanaka et al., 2001; Tanaka et al., 2007). The acr4-2 mutation has been described previously(Gifford et al., 2003). The ARF12 and ARF21 reporter lines were provided by Dolf Weijers. The reporter line N9185 was obtained from the Haseloff collection of enhancer trap lines(http://www.plantsci.cam.ac.uk/Haseloff/construction/GAL4Frame.html). The marker lines pATML1::GFP-ATML1 and pACR4::H2B-YFP have been described previously (Gifford et al.,2003).

To propagate zou mutants, seeds were sterilized and germinated on sterile tissue culture medium comprising 0.5 MS salts (Duchefa), 0.3% sucrose,1% agar. Seedlings were transferred to soil about 10 days after germination,when first leaves were easily visible.

Protein sequences similar to ZOU were retrieved using the BLASTP program(http://www.ncbi.nlm.nih.gov/blast/Blast.cgi)to search plant genome databases(http://www.plantgdb.org/). The Selaginella moellendorffi sequence was retrieved using the TBLASTN program to query the Selaginella DNA sequence database(http://selaginella.genomics.purdue.edu/cgi-bin/blast_tmpl_s.cgi). We used the software NetGene2(http://www.cbs.dtu.dk/services/NetGene2/)(Brunak et al., 1991; Hebsgaard et al., 1996) and comparisons with the ZOU sequence to predict intron/exon boundaries in the Selaginella sequence. The protein sequence alignments were generated using the programs T-Coffee and ClusalW2 available at http://www.ebi.ac.uk/t-coffee/.

Southern analysis showed that zou-1D plants carried a single pSKI074 T DNA insertion (data not shown). The T DNA contained a selectable marker (kanamycin resistance) that co-segregated with zou-1D (data not shown), suggesting that the insertion was the cause of the mutation. The sequences flanking the pSKI074 T-DNA insertion at zou-1D were isolated using the plasmid rescue technique described previously(Weigel et al., 2000). We confirmed the predicted intron-exon structure of ZOU by isolating a cDNA from silique tissue and determining its sequence, which matched that in databases (Swissprot Accession Number, NM 103864). Total RNA was extracted using Trizol Reagent (Invitrogen) and RNeasy columns (Qiagen) according to manufacturer's instructions. First-strand cDNA was prepared from 0.5 μg total RNA primed with oligo-dT primer using ImProm-II reverse transcription system (Promega) according to the manufacturer's instructions. The ZOU cDNA was amplified by PCR using primers ZOU-F 5′-GGCTCTAGATGACTAATGCTCAAG and ZOU-R 5′-CAGGTCGACAACTCAAACCGAAGC. The resulting product was cloned in the plasmid vector pGEM-T (Promega) to generate clone pSY3.

To generate the 35S::ZOU construct, the ZOU cDNA was excised from pSY3 by digestion with XbaI and SalI, and subcloned into the binary vector pFP101 (generous gift from Francois Parcy)under control of CaMV 35S promoter to generate clone pSY5. To generate the ZOU::ZOU-GFP reporter, the Gateway modified primers ZOUGWF(5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTGCCTGACCATAACAACCTATATCTC) and ZOUWGR(5′-GGGGACCACTTTGTACAAGAAAGCTGGGTTAGAGATGAAAAATATAACACCAGTTC) were used to amplify the ZOU genomic region from plant DNA (Col-0 ecotype) by PCR with the hi-fidelity Phu DNA polymerase (NEB). The PCR product was cloned into the Gateway entry vector pDONR207 by recombination with BP enzyme mix (Invitrogen) to create clone pSY13. The clone pSY13 was then recombined with the Gateway compatible binary vector pMDC111(Curtis and Grossniklaus,2003) to create clone pSY14 encoding an in-frame fusion of GFP to the C terminus of the ZOU protein. To create the ZOU::H2B-YFPreporter, a 1.6 kb region upstream of the ZOU ATG codon was amplified by PCR (as above) using primers ZOUSALIF (5′-GTCGACTGTGGTGGCATAATACGA)and ZOUSALIR (5′-GTCGACTGCTCATTTTACCCTTTT). The resulting product was subcloned into the vector pSC-B (Stratagene), excised by digestion with SalI and cloned into the binary vector pMD4(Gifford et al., 2003)containing the H2B-YFP fusion.

RT-PCR analysis of gene expression was performed as described previously(Chanvivattana et al., 2004)using the following primers: ZOU ZOUD3F,5′-GCTGACTATCTGTGGGAATG; ZOUD3R, 5′-AACTCGGATTTACCTGTGCT; EiF4A EiF4AF, 5′-TTCGCTCTTCTCTTTGCTCTC; and EiF4A-R,GAACTCATCTTGTCCCTCAAGTA. In situ hybridization analysis was as described previously (Chanvivattana et al.,2004), with the exception that plant tissue was wax embedded using an automated processor (Leica Tissue Processor TP1050) and embedding centre(Leica EG 1160). The ZOU probes were made from plasmid pSY3. To make the ALE1 probe, a region of ALE1 was amplified from silique cDNA using primers ALE5-2 (5′-TGAAACTAATGACAACATACACTCCC) and ALE3-2(5′-ACATATCACGATACTTCCAAAAACTGC). The resulting product was subcloned into pGEM-T vector (Promega). Plasmids were linearized by digestion with appropriate restriction enzyme, and RNA probes made using T7 or SP6 RNA polymerase as described previously(Chanvivattana et al.,2004).

For real-time PCR measurements, RNA was extracted from seedlings and siliques using Trizol Reagent (Invitrogen) and RNeasy columns (Qiagen)according to manufacturers instruction's. First-strand cDNA was prepared from 0.5 μg total RNA primed with oligo-dT primer using ImProm-II reverse transcription system (Promega) according to manufacturer's instructions. PCR reactions were performed in triplicate and the products quantified using a Rotor gene RG-3000 real-time PCR machine and associated software (Corbett Research) to assay SYBR green fluorescence. PCR reactions were made using SYBR green Jump-Start mix (Sigma). ZOU was amplified using primers ZOUD3F and ZOUD3R (above). ALE1 was amplified using primers ALE1F(5′-CTTCTCAGGCCAAGAAACTC) and ALE1R (5′-TTTGCCAGACTTGTTGAGGA). AtSUC5 was amplified using primers AtSUC5-F(5′-ATCGAAGAAACTTTACGACCAAGG) and AtSUC5-R(5′-TTAACGCTAAGACTCCACTAACC). Results were normalized using the EiF4A gene primers described above.

The expression of GFP reporter gene constructs was monitored by fluorescence microscopy using an Olympus Fluoview confocal microscope. For DIC microscopy, seeds were cleared overnight in modified Hoyer's solution (chloral hydrate:water:glycerol in proportions 8 g:2 ml:1 ml) and examined using a Nikon Eclipse E600 microscope. For thin sections, seeds were submerged overnight in 30 mM PIPES (pH 7) solution containing 5% glutaraldehyde and 0.1%(w/v) Triton-X-100. Tissue was rinsed several times in 30 mM PIPES and dehydrated using an ethanol series. LR White hard grade resin (London Resin Company, Reading, UK) was infiltrated into tissue over a period of 5 days before embedding. Ultramicrotome-cut sections (1 μm) were stained with Toluidine Blue and visualized using the above Nikon microscope. For transmission electron microscopy (TEM), seeds were fixed and embedded as above, except that they were stained with 1% aqueous osmium tetroxide prior to embedding. Ultrathin sections were cut using a diamond knife, stained with 2%aqueous uranyl acetate, and viewed on a Philips CM 120 transmission electron microscope. Cryo-scanning electron microscopy studies were carried out using a Hitachi S-4700 scanning electron microscope (Hitachi High Technologies,Wokingham, UK).

For double mutant construction, zou-4 mutants were crossed as females to acr4-2, ale1-1 or ale2-1 homozygotes (all alleles are in Colombia background, ale2-1 mutants are female sterile). In all cases, zou-4 homozygotes were pre-selected in the resulting F2 generation by only germinating the shrivelled (zou) seeds. To confirm that these plants were zou-4 homozygotes, we also tested that their F3 progeny was 100% resistant to sulphadiazine antibiotic (the T-DNA allele at zou-4 confers resistance). The zou4 acr4-1 mutant was identified from F2 plants whose immature F3 seeds were round (phenotype maternally determined by acr4-1 mothers). The ale1-1 zou-4double mutant was identified by a PCR-based genotyping assay using primers ALE1GenoF (5′-CGTGCTAGAATAGACGAAGG), ALE1GenoR(5′-CGTGGTGGAGATGGCAG) and ALE1Ds (5′-CCGTTTTGTATATCCCGTTTCCGT). The ALE1+ allele produced a 388 bp fragment, whereas ale1-1harbours a Ds transposon insertion and produces a 300 bp fragment(Tanaka et al., 2001). Because ale2 mutants are sterile, we first identified zou-4/zou-4 ale2-1/+ F2 individuals using primers ALE2F(5′-AGGAACGCTTGATTGGGATG) and ALE2R (5′-GAAGTCAGCAGAGTCTGGTA) (the ale2-1 mutation introduces a novel XhoI site within the amplified ALE2 fragment). These plants gave rise to F3 seeds that were germinated in sterile tissue culture and segregated about one-quarter of seedlings with severely defective cotyledons. We confirmed that these were zou-4 ale2-1 double mutants by PCR-based genotyping.

We identified the ZOU gene serendipitously, from a gain-of-function mutation, zou-1D, which we isolated by activation-tagging mutagenesis (Weigel et al., 2000) (see Materials and methods; see Fig. S1 in the supplementary material). The zou-1D mutation harboured a T-DNA insertion in the promoter of At1g49770, 119 nucleotides upstream of the ATG start codon (Fig. 1A). The insertion caused ectopic expression of At1g49770 in leaves, but did not affect the expression of At1g49760, the other gene neighbouring the insertion (Fig. 1B). The mis-expression of At1g49770 caused leaf curling(see Fig. S1 in the supplementary material). However, the biological relevance of this phenotype is unclear as ZOU is not normally expressed in leaves and the leaf curling is not obviously related to the normal function of ZOU. In this study we therefore concentrate on the analysis of loss-of-function mutations.

The ZOU gene is predicted to encode a transcription factor of the basic helix-loop-helix (bHLH) class, designated bHLH95 in previous analyses of the Arabidopsis genome (Heim et al., 2003; Toledo-Ortiz et al., 2003). Genome database searches revealed that the ZOU protein is widely conserved in plants (see Fig. S2 in the supplementary material):homologues were found both in other dicotyledonous species, including grape(Vitis vinifera) and barrel medic (Medicago truncatula) and also in more distantly related angiosperms (e.g. rice, a monocot);furthermore, ZOU homologues were also found outside flowering plants in the gymnosperms [Sitka spruce (Picea sitchensis)] and club mosses(Selaginella moellendorffii). Protein sequence alignments showed that two regions of the ZOU protein were well conserved: the bHLH domain (residues 66-124) and a region of unknown function near the C terminus of the protein(residues 215-301, see alignment in Fig. S2 in the supplementary material). It is likely that these proteins are ZOU orthologues as each is more similar to ZOU than to any other Arabidopsis protein. Using these criteria, no clear ZOU orthologues were identified in the bryophyte Physcomitrella patens or in the green alga Chlamydomonas reinhardtii. The ZOU gene is therefore ancient and is likely to have arisen early in vascular plant evolution. In addition, the Arabidopsis ZOU protein is more similar to homologues in other plant species than to any of the other Arabidopsis bHLH proteins [estimated to be between 132 and 146 in number (Heim et al., 2003; Toledo-Ortiz et al., 2003)],suggesting that it is a unique gene in Arabidopsis.

A previous study of Arabidopsis bHLH genes suggested that ZOU (there termed bHLH95) expression was specific to flowers and/or siliques (Heim et al., 2003). Quantitative RT-PCR (Fig. 1C)confirmed and extended this: ZOU was not expressed in whole seedlings, rosette leaves or unopened flower buds, but was detected at a low level in opened flowers and more strongly in siliques, suggesting that it normally functions in siliques or seed development. To determine the loss-of-function phenotype, we obtained three independent alleles (zou-2,zou-3 and zou-4) from collections of T-DNA insertion lines(Fig. 1A). All three alleles conferred very similar phenotypes, with zou-2 being most severe,consistent with the location of the T-DNA insertion in the first exon. Plants heterozygous for zou were normal; however, when self-pollinated about one quarter of their seeds (see Table 1) were mis-shapen and shrivelled at maturity(Fig. 2A,B). In zou-2homozygotes, all seeds were shrivelled, but when they were crossed either as male or female parent to wild type, the resulting seeds were normal(Table 1). Together, these observations indicated that zou-2-4 are recessive loss-of-function mutations with normal transmission through male and female gametophytes and that ZOU acts zygotically (rather than maternally) to control seed development. The seed shrivelling phenotype co-segregated with antibiotic resistance conferred by the T DNA insertion at zou-2 (data not shown)and could be complemented by transformation with a genomic ZOU clone(Fig. 2C; Table 1), confirming that it was caused by loss of ZOU function. We renamed the At1g4977/bHLH95 gene ZHOUPI (Chinese for `shrivelled') in reference to the seed phenotype.

Mature seeds of zou mutants contained smaller embryos than wild type (Fig. 2D), suggesting that the shrivelling of zou seeds occurs because the embryo fails to fill the seed. However, zou seeds were viable (see below), indicating that zou embryos do not undergo premature developmental arrest. To determine when zou mutations first affect embryo development,developing seeds were cleared and examined using differential interference contrast (DIC) microscopy. Up until the early heart stage of development, zou embryos appeared normal and the surrounding endosperm became cellularized around early heart stage, as in wild type(Fig. 2E,F). Subsequently, zou embryos failed to elongate and were shorter and wider than wild-type torpedo and later stage embryos. Whereas in wild type the cotyledons elongate and bend over, so that a large embryo fills the seed, in zouembryos the cotyledons remained short and stumpy, and were usually more widely separated (Fig. 2G,H). The failure of zou embryos to elongate may be due to their abnormal adhesion to the surrounding endosperm, which physically impedes growth. In wild-type seeds, the endosperm in the region surrounding the embryo (ESR)separates from the embryo from the heart stage of embryogenesis onwards; at the torpedo stage of development, little or no endosperm remains between the cotyledons (Fig. 2G). By contrast, zou embryos continue to adhere to endosperm throughout embryogenesis and considerable amounts of endosperm are present between cotyledons (Fig. 2H). To characterize the adhesion of endosperm to embryos more clearly, we examined thin sections of resin embedded seeds using light microscopy. This confirmed that in zou seeds the endosperm remains stuck to the embryo over its entire surface (Fig. 2I,J). Similar, but less severe, defects in ale1 mutants(Tanaka et al., 2001) have been attributed to impaired cuticle formation during embryogenesis. To examine surface development in zou embryos, seeds were embedded in resin,stained with osmium tetroxide and ultra-thin sections examined using transmission electron microscopy (TEM). Wild-type torpedo stage embryos were covered by a cell wall with a continuous electron dense cuticularized layer(Jeffree, 2006) (CL). An overlying translucent cuticle proper(Jeffree, 2006) (CP) was also present, although visualization of this at high resolution in seed tissue is challenging. We did not observe any obvious discontinuities in the CL of zou mutant embryos at the same developmental stage, but no CP was apparent, and the cuticularized cell wall was extensively fused to endosperm cell wall (Fig. 2K,L).

The separation of embryo from endosperm during angiosperm embryogenesis correlates with breakdown of the endosperm cells surrounding the developing embryo (Briggs, 1993). As this separation does not occur efficiently in zou mutants, we examined whether the endosperm is more persistent in zou seeds than in wild type. After imbibition of mature wild-type seeds, the embryo could easily be separated from the single layer of residual endosperm cells by extrusion. By contrast, in zou mutant seeds, the embryo and endosperm adhered strongly. Furthermore, the chalazal pole of the endosperm, which was not invaded by the embryo, formed a sac-like structure. When the embryo was cut away, a considerable quantity of abnormal paste-like tissue was extruded from the endosperm cavity (see Fig. S3 in the supplementary material).

The shrivelled zou seeds were viable and germinated to produce seedlings that survived when grown in conditions of high humidity, but desiccated and died when grown in low humidity. This phenotype is commonly found in mutants with defects in cuticle or epidermis that impair water retention. We therefore stained whole seedlings with Toluidine Blue, as this has been shown to provide a rapid test for cuticle and/or epidermal defects(Tanaka et al., 2004). Whereas wild-type plants were unstained, in zou mutants the hypocotyl and the cotyledons, but not the leaves, stained strongly(Fig. 3A,B). This suggested that the defects in zou mutants were restricted to organs that initiate during embryogenesis (i.e. hypocotyls and cotyledons, but not leaves), consistent with the seed-specific expression of ZOU(Fig. 1C; Fig. 4). Furthermore, when zou mutants were grown in tissue culture and transferred to soil after the first leaves had initiated, the mutants recovered and gave rise to normal plants, again suggesting that ZOU did not affect post-embryonic organ development.

Epidermal defects in zou seedlings were examined using scanning electron microscopy (cryo-SEM). The leaves of zou mutants were normal(Fig. 3C). The epidermal surfaces of zou cotyledons had normal pavement and stomatal cell types (Fig. 3D,E); however,large tears were frequently visible in the epidermis through which the underlying mesophyll cells protruded (Fig. 3C,F). As the protoderm appeared continuous in zouembryos (Fig. 2F,H,J), it is likely that the tears developed after germination, when cotyledon expansion occurred, owing to epidermal defects. Consistent with this, in less severely affected zou mutants, small holes were visible between cells, as if cell adhesion were compromised (Fig. 3G). To analyse post-germination cotyledon surface structure, TEM analysis was carried out on expanded wild-type and zou cotyledons 7 days post germination. Wild-type cotyledons were covered in a continuous cell wall with a pronounced CL and a clearly demarcated overlying CP. The CP was demarcated by a fine outer electron-dense line, possibly indicating an outer proteinaceous layer (Fig. 3J). In zou cotyledons, variability in thickness and discontinuities in the CP were evident, and the outer electron dense layer overlying the CP was more pronounced (Fig. 3K). This was not simply a consequence of tears, as abnormalities were visible in regions where the epidermal sheet was intact. In addition, remnants of endosperm were visible adhering to zou, but not to wild-type,cotyledons (Fig. 3L).

We combined zou mutations with other mutations implicated in epidermal development, including acr4, ale1 and ale2 (see Materials and methods). The zou-4 mutation was epistatic to ale1-1 in double mutants, suggesting that ZOU and ALE1 might act in the same pathway (data not shown). The zou-4 acr4-2 seedlings had more severely defective cotyledons than either single mutant, suggesting that ZOU and ACR4 may be involved in a common developmental process (Fig. 3H). The zou-4 ale2-1 seedlings also had more severely defective cotyledons than either single mutant(Fig. 3I), consistent with previous observations that ale1 mutations enhance ale2(Tanaka et al., 2007).

To determine where ZOU was expressed, we localized ZOUmRNA by in situ hybridization to sections of developing seeds. We detected strong expression in the endosperm in the ESR in seeds containing embryos up to the torpedo stage, after which expression declined(Fig. 4A-C). No signal was detected when sense control probes were hybridized, confirming that the signal with antisense probes was specific for ZOU(Fig. 4D). Strikingly, we did not detect expression either in the embryo itself or in any of the maternal tissues in the seeds or siliques. The effects of ZOU on the embryonic epidermis are therefore non cell autonomous. To further characterize ZOU expression, we made two reporter gene constructs: a gene fusion in which the entire ZOU genomic region (i.e. upstream promoter and intragenic sequences) was fused in frame to GFP (ZOU::ZOU-GFP); and a fusion of the ZOU upstream promoter sequences to a HISTONE 2B-YFP fusion protein (Boisnard-Lorig et al.,2001) (ZOU::H2B-YFP). Both constructs showed similar expression patterns. The ZOU::ZOU-GFP construct complemented zou-2 mutants (Fig. 2Cand Table 1), confirming that the ZOU-GFP protein fusion retained ZOU+ activity and that the construct was expressed correctly. The ZOU-GFP fusion protein was nuclear-localized, consistent with the predicted function of ZOU as a transcription factor, and was expressed in the ESR but not in the embryo(Fig. 4E). The non cell-autonomous effects of ZOU on the embryonic epidermis are therefore not due to the ZOU protein moving from endosperm to embryo. The ZOU-GFP fusion protein expressed from ZOU::ZOU-GFP was below the limits of detection by confocal microscopy until the early heart stage, possibly owing to post-transcriptional regulation. For this reason, we characterized the early expression of ZOU using the ZOU::H2B-YFP construct. ZOU was not expressed in the ovule prior to fertilization(Fig. 4F) but was strongly expressed in the central cell after fertilization(Fig. 4G). This is consistent with Q-PCR data which showed that ZOU expression was undetectable in unopened flower buds, but was detectable in opened flowers shortly after the point of fertilization (Fig. 1C). It was expressed uniformly throughout the endosperm during the early stages of its development when it comprised two to eight nuclei(Fig. 4H,I), but then became expressed more strongly at the micropylar pole during the 12-16 nuclei stage(Fig. 4J) and the 24-28 nuclei stages (Fig. 4K); it was largely restricted to the ESR by the 44-48 cell stage[Fig. 4L, stages as defined previously (Boisnard-Lorig et al.,2001)].

ZOU showed a very similar expression pattern to ALE1. The two genes have similar mutant phenotypes(Tanaka et al., 2001) and zou mutants are epistatic to ale1 mutants, suggesting that they might act in a common pathway. To test whether ZOU regulates ALE1, consistent with the likely role of ZOU as a transcription factor, we localized ALE1 expression in young siliques of wild-type and zou-2 homozygotes. ALE1 was strongly expressed in the ESR of the endosperm but not in the embryos of wild-type seeds(Fig. 5A,B). In zou-2mutants, ALE1 expression was barely detectable at globular and heart stages (Fig. 5C,D), indicating that ZOU regulates ALE1 expression. To test whether ZOU had a general role in regulating ESR identity, or rather was specific for genes such as ALE1 (which have a role in embryo development), we analyzed the expression of AtSUC5, a sucrose transporter that is also expressed specifically in the ESR(Baud et al., 2005), using quantitative RT-PCR. Unlike ALE1, which showed an approximately fivefold decrease in expression, there was no significant change in AtSUC5 expression in zou mutants (see Fig. S4 in the supplementary material). Likewise, the expression of reporters for the ESR-specific AUXIN RESPONSE FACTOR 12 (ARF12) and ARF21 genes (D. Weijers, personal communication) was unaffected in zou mutant seeds (data not shown), confirming that ZOU does not regulate all ESR-specific genes. However, the enhancer trap line N9185,which is expressed in wild-type ESR(Ingouff et al., 2005) was not expressed in zou-4 (Fig. 5E,F), suggesting that the gene responsible for the N9185 pattern may also be a target of ZOU regulation. The enhancer trap T-DNA in N9185 is located on chromosome 4 (data not shown), and is therefore not associated with ALE1 (At1g62340).

Because zou mutant seedlings had epidermal defects, we also introduced reporters for ACR4 and ATML1, which show epidermal-specific expression. Both reporters were expressed normally in zou-4 mutant embryos (Fig. 5G-J), consistent with the fact that protoderm specification appears largely normal in cleared and sectioned zou seeds.

The fact that ZOU is needed for ALE1 expression in seeds raised the possibility that the curled leaves of zou-1D mutants could be a result of ALE1 being activated by ZOU ectopically in leaves. To test this, we analysed ALE1 and ZOU expression in seedlings by quantitative RT-PCR. Consistent with our RT-PCR results(Fig. 1B), ZOUexpression was not detectable in wild-type seedlings, whereas in zou-1D/+ seedlings it was readily detectable at about six times the level in wild-type siliques. No ALE1 expression was detectable in wild-type or zou-1D seedlings (see Fig. S5 in the supplementary material). Therefore, ZOU expression is not sufficient to activate ALE1 outside of the seed and the effects of ZOUmis-expression may be unrelated to the normal function of ZOU in seeds.

In this paper, we identify ZOU as a transcription factor whose expression is entirely restricted to the developing Arabidopsis endosperm during seed formation, but whose function mediates the normal development of neighbouring embryonic tissues. The potential non cell-autonomous role of the angiosperm endosperm as a source of developmental signals for the young embryo has long been a source of speculation. The observation that in vitro embryogenesis from cell cultures is often accompanied by the expression of endosperm-specific genes in non-embryogenic cells, led to the idea that the`nurse-cell'-like capacities of these cells could be analogous to the developmentally important roles of the endosperm during zygotic embryo development (Magnard et al.,2000; Massonneau et al.,2005; van Hengel et al.,1998; Wiweger et al.,2003). However, the first concrete evidence that the endosperm might play a key role in Arabidopsis embryo development, distinct from its role in embryo nutrition, came only recently, with the characterization of the ESR gene ALE1(Tanaka et al., 2001). We show that ZOU regulates ALE1 expression in endosperm, consistent with the similar expression patterns of the two genes. However, the zou phenotype is not simply a consequence of a loss of ALE1 activity because ale1 mutants are much less severe than zou mutants. In our growth conditions, ale1 mutant seeds did not shrivel, they produced normal sized embryos and the epidermal morphology of their seedlings was only mildly affected. It is likely that ZOU regulates many targets additional to ALE1, as is supported by the loss of the N9185 reporter gene expression in zou mutants. Previous analyses of the Arabidopsis bHLH family indicated that all residues crucial for DNA-binding activity and dimerization are conserved in the ZOU bHLH and that it is likely to bind the G box sequence CACGTG(Heim et al., 2003; Toledo-Ortiz et al., 2003). Because mis-expression of ZOU in leaves is insufficient to activate ALE1, ZOU probably requires additional factors for activity. One possibility is that it acts as a heterodimer with another bHLH member, as is the case for several plant bHLH proteins(Heim et al., 2003; Toledo-Ortiz et al., 2003). Because ZOU is expressed in rather few cells in seeds (i.e. the ESR of endosperm), it is technically difficult to demonstrate direct binding in vivo using techniques such as chromatin immunoprecipitation.

The ESR of the angiosperm endosperm is distinct structurally, and also in terms of gene expression, from the rest of the endosperm(Berger, 2003; Olsen, 2004). In both cereals and Arabidopsis, where the first few divisions of the central cell give rise to a free nuclear syncitium, cellularization is initiated in the ESR. The ESR endosperm is also densely cytoplasmic and is closely associated with the developing embryo at early stages. This association is probably key for provision of nutrients to the young embryo, a supposition that is supported by the ESR-specific expression of the sucrose transporter AtSUC5,which is required for normal embryo growth in Arabidopsis(Baud et al., 2005). ZOU is the first transcription factor to be identified with an ESR-specific expression pattern, raising the possibility that it specifies the identity of the ESR region of the endosperm. We feel that this is unlikely for three reasons:first, the early morphology and development of zou endosperm,including the ESR, is indistinguishable from that of wild type; second, the expression of some ESR genes, such as AtSUC5, ARF12 and ARF21, is not affected in zou mutants; finally, the expression of ZOU does not become wholly restricted to the ESR until the early heart stage. Therefore, rather than specifying the identity of the ESR region, it seems more likely that ZOU has more specific roles in endosperm and or embryo development.

Although zou mutants form a defective epidermis, it is unlikely that ZOU is absolutely required for epidermal specification, as the expression of epidermal specific marker genes such as ATML1 and ACR4 appears normal in zou embryos. However, we have shown that the translucent cuticle-proper, which usually prevents cotyledon dehydration, is discontinuous in zou mutants. These discontinuities are not simply a result of the tears and holes that frequently develop in the epidermal surfaces of zou mutants, as they can be found in cuticle that overlies intact epidermal sheets. Thus, one function of ZOU is to allow formation of a normal cuticular membrane. It is possible that the cuticular defects are an indirect consequence of the persistent abnormal juxtaposition of living endosperm with the embryo. However, a very similar endosperm adhesion phenotype results when two embryo-specific receptor-like kinases, encoded by the GSO1 and GSO2 genes, are disrupted,indicating that endosperm adhesion is also controlled by the embryo and may be a result of cuticle defects (Tsuwamoto et al., 2008). We therefore suggest that ZOU regulates signalling between the endosperm ESR and the embryo that promotes normal cuticle deposition and is probably mediated by receptor-like kinases such as GSO1 and GSO2 in the embryo. Such signalling is probably two-way: ZOUexpression becomes gradually restricted to the ESR as endosperm development progresses. The maintenance of ZOU expression in the endosperm cells directly in contact with the embryo may therefore be the result of signalling between the embryonic epidermis and the endosperm. Whether cuticle deposition is entirely mediated by the embryonic epidermal cells, or involves secreted compounds originating from the endosperm, is, as yet, unclear.

In Arabidopsis and many other dicotyledonous plants, the cellularized endosperm progressively degenerates as it is invaded by the expanding embryo, so that only a single layer - the aleurone - remains at seed maturity. By contrast, in mature zou seeds, much more endosperm persists, including non-aleurone tissues. Therefore, a likely second role of ZOU is to promote the expression of genes involved in degradation of endosperm cells around the expanding embryo. The ZOU target ALE1 encodes a subtilisin-like serine protease that has been proposed to process ligands involved in epidermal specification/function(Watanabe et al., 2004). However, it is also possible that ALE1 acts instead in the separation of the developing endosperm from the embryonic epidermis; for example, by promoting endosperm autolysis.

ZOU is therefore the first transcription factor to be identified that mediates two poorly understood processes in seed development - separation of the embryo from the endosperm and breakdown of the endosperm. Determining the targets of ZOU; for example, by transcriptional profiling of mutant seeds,should help identify the key factors that mediate these processes.

It is striking that ZOU is so widely conserved in plants. Within angiosperms, it is likely that the role of ZOU in the ESR is conserved as the rice ZOU homologue also shows seed-specific expression (Li et al., 2006). However, ZOU is also found both in seed plants that lack an endosperm(e.g. Picea sitchensis, a gymnosperm) and also in more basal vascular plant groups, such as the club mosses, which lack seeds altogether. A common feature of all these groups of land plants is that they have a specialized epidermis with cuticle, and that their embryos enlarge by invading and digesting surrounding maternal nutritive tissues (see Fig. S6 in the supplementary material). In gymnosperms, the female gametophyte proliferates to form a nutrient-rich tissue prior to fertilization, and the growth of the embryo post-fertilization is invasive (see Fig. S6B in the supplementary material). Although club mosses such as Selaginella lack a seed, the embryo is similarly thrust into a nutrient-rich megagametophyte by a suspensor(see Fig. S6C in the supplementary material). It is therefore possible that the ancestral role of ZOU is to promote the separation of the embryo from surrounding gametophytic tissues, and the breakdown of these tissues,similar to its role in angiosperm endosperm. This would also be consistent with the current consensus that the endosperm is evolutionarily homologous to the gymnosperm female gametophyte [see Baroux et al.(Baroux et al., 2002) for recent discussion]. To test this further, it will be necessary to determine whether ZOU is also expressed in the ESR of the female gametophyte in these groups.

While this article was under review, Kondou et al.(Kondou et al., 2008)published on RETARDED GROWTH OF EMBRYO1. The genes ZHOUPI and RETARDED GROWTH OF EMBRYO1 both correspond to At1g49770.

J.G. and S.Y. were funded by a BBSRC grant in the Integrated epigenetics initiative. G.I. was funded by an RCUK fellowship. We thank Karen Halliday for use of her real-time PCR machine, Kim Johnson for experimental advice, Fred Berger and Dolf Weijers for reporter lines, and Christine Stock for providing the image in Fig. 5E.