Endosperm breakdown in Arabidopsis requires heterodimers of the basic helix-loop-helix proteins ZHOUPI and INDUCER OF CBP EXPRESSION 1

Angiosperm seed development necessitates coordinated growth of the two products of fertilization, i.e. the endosperm and the embryo, within the maternally derived seed coat. In wild-type Arabidopsis, the syncytial endosperm expands soon after fertilization and then cellularizes (Garcia et al., 2005; Ingram, 2010; Sørensen et al., 2002). The space occupied by the endosperm after its cellularization is subsequently appropriated by the developing embryo as it expands, so that by the end of seed development only a single layer of endosperm cells remains. The process of endosperm breakdown is regulated by a unique and highly evolutionarily conserved basic helix-loop-helix (bHLH) transcription factor, ZHOUPI (ZOU). In zou mutants, the endosperm persists throughout seed development, significantly impeding the growth of the embryo and giving rise to a collapsed seed phenotype upon desiccation (Xing et al., 2013; Yang et al., 2008).

In addition to regulating the breakdown of the endosperm, ZOU regulates the expression of the subtilisin serine protease ALE1 in the endosperm, which in turn acts with two embryo-expressed receptor kinases, GASSHO1 and GASSHO2 (GSO1 and GSO2), to regulate embryonic cuticle biogenesis. Mutants in ALE1, GSO1 and GSO2 have no measurable defect in endosperm breakdown but produce abnormal embryonic surfaces (Tanaka et al., 2001; Xing et al., 2013; Yang et al., 2008).

bHLH transcription factors are known to act combinatorially as homo- or heterodimers to regulate developmental processes in both animals and plants (Jones, 2004; Massari and Murre, 2000; Pires and Dolan, 2010). To understand more about ZOU function, we investigated whether this unique protein might act in combination with other bHLH transcription factors in order to carry out one or both of its roles in the developing seed.

To identify potential protein partners of full-length ZOU protein, a yeast two-hybrid screen was carried out using a split ubiquitin-based system in which the bait protein is presented as a membrane-anchored fusion (Möckli et al., 2007). Two independent cDNA clones encoding similar truncated versions of the bHLH protein INDUCER OF CBP EXPRESSION 1 (ICE1, also known as SCREAM; AT3g26744) were obtained in a screen of 1.2 million clones (Fig. 1A; see Materials and Methods). Both clones encoded the bHLH domain-containing C-terminal region of the ICE1 protein (ICE1-C). ICE1-C was confirmed as a true interactor of ZOU by retransformation with the ICE1-C-containing prey vector and either the ZOU-containing bait vector or an empty bait vector, with growth on selective media only being observed in the former case (Fig. 1B).

Further constructions encoding full-length ICE1 (ICE1-FL) or ICE1-C as bait or prey fusions were used to test interactions with full-length ZOU (ZOU-FL) or with a truncated version lacking the bHLH domain (ZOU-C). Results are summarized in Fig. 1B. The ICE1-FL protein showed only a weak interaction with ZOU-FL in yeast. Although the ZOU-FL construct interacted strongly with ICE1-C, ZOU-C showed no interaction, suggesting that, as expected for this protein family, the bHLH domain-containing regions of the two proteins mediate their interaction (Jones, 2004). Homodimerization of both ZOU-FL and ICE1-C was detected in yeast.

ZOU expression is strictly localized to the developing endosperm (Yang et al., 2008). In situ hybridizations were carried out in developing seeds and, consistent with in silico data (Le et al., 2010), expression of ICE1 was detected in all three seed tissues, the endosperm, the embryo and the testa. Expression in the embryo-surrounding endosperm was particularly strong (Fig. 1C; supplementary material Fig. S1). ZOU and ICE1 are therefore co-expressed in the endosperm during seed development.

ICE1 has well characterized roles in regulating both tolerance to cold stress (Chinnusamy et al., 2003; Lee et al., 2005; Miura et al., 2007) and stomatal development, in which the class III bHLH protein ICE1 and its close homolog SCREAM2 (SCRM2, also known as ICE2) have been shown to interact with the class Ia bHLH proteins SPEECHLESS, MUTE and FAMA (Kanaoka et al., 2008). We obtained the previously characterized null ICE1 allele ice1-2 (Kanaoka et al., 2008) and observed that homozygous plants produced shrivelled seeds (Fig. 2A-C). This phenotype could be complemented by an ICE1:GFP protein fusion expressed under the ICE1 promoter (supplementary material Fig. S2), confirming that this phenotype is caused by loss of function of ICE1. In resin-embedded silique samples, we observed that, like zou-4 mutants, ice1-2 mutant seeds have a persistent endosperm and much reduced embryo (Fig. 2E-G). To test for seedling cuticle defects, ice1-2, zou-4 and Col-0 seedlings were treated with the hydrophilic dye Toluidine Blue. Both ice1-2 and zou-4 seedlings showed similar, high levels of Toluidine Blue uptake compared with Col-0 seedlings (Fig. 2I-L). Thus, ICE1, like ZOU, is involved both in endosperm degradation and seedling surface formation.

The expression of five previously identified targets of ZOU activity (Xing et al., 2013), including ALE1, was investigated in staged zou-4, ice1-2 and Col-0 siliques containing embryos between the late globular and late heart stage of development. The transcription of all five genes was found to be lost in both the zou-4 and ice1-2 mutant background, showing that, consistent with their similar mutant phenotypes, ICE1 and ZOU regulate common targets (Fig. 3A; supplementary material Fig. S3). To test whether, like ICE1, ZOU and ALE1 (Yang et al., 2008), the other potential targets were strongly expressed in the endosperm, in situ hybridizations were carried out. Strong endosperm expression was detected in each case (Fig. 3C-E; supplementary material Fig. S3) and was not detected in either zou-4 or ice1-2 mutant backgrounds (Fig. 3F-H and supplementary material Fig. S3 show results for zou-4), consistent with the regulation of expression of these genes by ZOU and ICE1 in the endosperm.

To investigate the genetic relationship between ZOU and ICE1 during seed development, double mutants were generated between zou-4 and ice1-2. Double-mutant plants were viable and their seed and seedling permeability phenotypes were indistinguishable from those of either single mutant (Fig. 2D,H,I). To test whether ZOU and ICE1 act sequentially in a transcriptional cascade, their expression levels were compared in wild-type, zou-4 and ice1-2 mutant backgrounds. No significant misregulation of ICE1 was observed in zou-4 mutants, or vice versa (Fig. 3B). The independence of the phenotypes of ice1-2 and zou-4 is therefore likely to be due to a non-redundant functional role for the ICE1/ZOU heterodimer in seed development.

In addition to the recessive loss-of-function allele ice1-2, previous studies have described ice1-1/Scrm-D, which is caused by a point mutation giving rise to a single amino acid substitution in the ICE1 protein (Chinnusamy et al., 2003; Kanaoka et al., 2008; Miura and Ohta, 2010). During stomatal development, this mutation causes a strong semi-dominant gain-of-function phenotype opposite to that observed in ice1-2 scrm2 loss-of-function mutants, so that Scrm-D plants produce stomata over most of their cotyledon and leaf surfaces (Kanaoka et al., 2008). Despite this phenotype, Scrm-D plants are fertile, although silique initiation is slow and erratic. Mature seeds of Scrm-D homozygous plants were indistinguishable from those of wild type (Fig. 4G). When homozygous Scrm-D plants were crossed reciprocally with ice1-2, the resulting seeds resembled those of wild type, consistent with the Scrm-D allele functionally compensating for the presence of the recessive ice1-2 allele in the heterozygous state during seed development.

To assess whether Scrm-D affects endosperm development, the structure of developing Scrm-D homozygous seeds was examined. Although the endosperm of heart stage Scrm-D homozygous seeds started to cellularize apparently normally (Fig. 4A,D), by the torpedo stage, when the majority of wild-type endosperm is cellularized, the endosperm of Scrm-D homozygous seeds is densely packed around the periphery of the embryo cavity and the cell walls are particularly difficult to distinguish [phenotype observed clearly in 13 of 17 Scrm-D seeds sectioned at this stage, and in no simultaneously processed wild-type seeds (N=15)] (Fig. 4B,E). Mature non-desiccated Scrm-D homozygous seeds showed a large gap between the cotyledons and the hypocotyl (17 out of 19 seeds sectioned), which was not apparent in wild-type seeds processed simultaneously (N=16) (Fig. 4C,F). The phenotype of Scrm-D homozygous seeds is suggestive of precocious breakdown of the endosperm cell structure, confirming that Scrm-D is a hypermorphic allele of ICE1 with respect to endosperm breakdown. Interestingly, homozygosity for Scrm-D did not significantly alleviate the small embryo phenotype of the zou-4 seed. In this background the endosperm was well cellularized and persistent at seed maturity, as in zou-4 mutants (Fig. 4H,I). Thus, the precocious endosperm breakdown phenotype of Scrm-D homozygous mutants is dependent upon the presence of functional ZOU protein within the developing seed.

Our results suggest that heterodimerization of ZOU and ICE1 is strictly necessary for target gene activation. The gene At3g08900 was selected as a likely direct target of ZOU/ICE1, since its expression in endosperm is more than 1000-fold downregulated in both zou-4 and ice1-2 mutant backgrounds (supplementary material Fig. S3). We tested the activation of the At3g08900 promoter by ZOU, ICE1 or SCRM-D and by combinations of ZOU with ICE or SCRM-D in a Physcomitrella patens protoplast system (Thévenin et al., 2012), and confirmed that only combinations of ZOU with ICE or SCRM-D could activate promoter activity to levels above background (supplementary material Fig. S3).

To eliminate the possibility that the formation of the ZOU/ICE1 heterodimer is necessary for protein localization, constructs expressing functional ZOU and ICE1 tagged with GFP under their native promoters (Yang et al., 2008) (supplementary material Fig. S2) were crossed into the reciprocal mutant background. Despite complementing the ice1-2 mutant seed phenotype and being clearly visible in stomatal guard cells (supplementary material Fig. S2), the ICE1-GFP fusion protein could not be clearly visualized in the developing endosperm in any background. However, nuclear ZOU-GFP protein was present in both wild-type and ice-2 mutant seeds, confirming that ICE1 is not necessary for ZOU localization in endosperm nuclei (Fig. 4J,K).

The roles of SPEECHLESS, MUTE, FAMA, ICE1 and SCRM2 in the regulation of sequential developmental decisions in stomatal development show similarities with bHLH-regulated cascades in animals controlling developmental processes, including vertebrate myogenesis and Drosophila neurogenesis (Serna, 2009). In each case, tissue-specific factors preferentially form heterodimers with structurally distinct and more widely expressed bHLH proteins (Jones, 2004; Massari and Murre, 2000; Murre et al., 1989). Here we show that another highly tissue-specific bHLH protein, ZOU, also forms heterodimers with ICE1 to regulate endosperm breakdown. As has previously been observed in interactions between tissue-specific and ubiquitous bHLH proteins in animal systems, we show that although both ICE1 and ZOU can form homodimers in yeast, ICE1/ZOU heterodimers are responsible for the biological activity of these proteins in the endosperm (Jones, 2004; Massari and Murre, 2000; Murre et al., 1989).

Although ZOU possesses a unique C-terminal domain, recent phylogenetic analyses have placed it in bHLH class Ib1, relatively close to the class Ia SPEECHLESS, MUTE and FAMA proteins (Pires and Dolan, 2010). In contrast to ZOU, which is not clearly conserved in bryophytes (Yang et al., 2008), genes encoding both ICE1-like class IIIb and class Ia bHLH proteins can clearly be distinguished in the bryophyte Physcomitrella, which, like angiosperms, produces true stomata in the epidermal cell layer of the sporophyte (Vatén and Bergmann, 2012), leading to the suggestion that the class Ia/IIIb partnership might be ancient, possibly coinciding with the rise of the bryophytes and potentially contributing to the elaboration of true stomata (MacAlister and Bergmann, 2011; Vatén and Bergmann, 2012; Vatén and Bergmann, 2013).

ZOU expression in Arabidopsis is strictly limited to the endosperm, which in angiosperms is thought to be the sexualized homolog of the megagametophyte in lower land plants. Neofunctionalization of ZOU might have arisen after restriction of its expression to the astomatous gametophyte generation in a lycophyte ancestor, permitting rapid sequence divergence of ZOU relative to its more widely expressed and promiscuous class IIIb partner (ancestral ICE1). Neofunctionalization after changes in expression patterns has been evoked to explain the evolution of tissue-specific bHLH proteins involved in neural development in animal lineages (Simionato et al., 2008; Vervoort and Ledent, 2001). Detailed analysis of the expression patterns of class IIIb, class Ia and, where present, class Ib genes in lower plant species should serve to test this hypothesis in future studies.

Mutant Arabidopsis thaliana lines used in this study have been published previously. All plant lines were plated on Murashige and Skoog (MS) media (with or without selection), stratified for 3 days at 4°C, and germinated in a Lemnagen growth cabinet under long days (16 hours light) at 21°C for 7-10 days before transferring to soil in identical growth room conditions.

Plant DNA was extracted using a rapid CTAB isolation technique as described (Stewart and Via, 1993). zou-4 genotyping was carried out as previously described (Tanaka et al., 2001; Xing et al., 2013; Yang et al., 2008). Genotyping for ice1-2 was performed using primers ICE T-DNA F and ICE T-DNA R for the wild-type allele and ICE T-DNA F with a SALK left border (LB) primer for the mutant allele (supplementary material Table S1).

Procedures for the quantification of Toluidine Blue uptake were as described (Xing et al., 2013). Etiolated seedlings were stratified for 3 days at 4°C, exposed to 4 hours of light in a Lemnagen growth cabinet at 21°C, and then plates were covered with two layers of metal foil before allowing seedlings to germinate and grow for 5 days. Coloration was carried out as previously described (Xing et al., 2013).

Resin embedding and Periodic Acid Schiff (PAS) staining procedures were carried out exactly as described previously (Xing et al., 2013).

Methods were as previously described (Yang et al., 2008). Antisense probes were transcribed from linearized pCR BluntII-TOPO vectors (Invitrogen) containing full-length ORFs amplified from silique cDNA with ICE 5′ YEAST and ICE 3′ YEAST BAIT, 1g71250cdsFor and 1g71250cdsRev, 2g43870cdsFor and 2g43870cdsRev, or At3g38000cdsFor and At3g38000cdsRev (supplementary material Table S1).

Protocols for staging of plant material and qRT-PCR analysis, including qRT-PCR primers for all genes except ICE1, have been described previously (Xing et al., 2013; Yang et al., 2008). Levels of ICE1 were detected using ICE1-Q-L and ICE1-Q-R (supplementary material Table S1).

Yeast two-hybrid experiments were carried out using the Dual Hunter split ubiquitin-based system from Dualsystems Biotech (Schlieren, Switzerland) (Möckli et al., 2007). All yeast protocols were performed exactly as described in the manufacturer’s instructions. Single bait and prey constructs were made by cloning PCR-amplified cDNA fragments flanked by in-frame SfiI sites into pCR BluntII-TOPO vector (Invitrogen), and then transferring SfiI-excised inserts into the appropriate yeast vector. Primers used for specific constructs are listed in supplementary material Table S1. A normalized cDNA library was generated commercially by Dualsystems Biotech from RNA isolated from a pool of developing siliques containing embryos ranging from the zygote to the mid-torpedo stage of development. RNA was isolated using the Spectrum Plant Total RNA Kit (Sigma). Library screening was carried out under stringent conditions on media lacking both adenine and histidine.

Moss culture, protoplast preparation, vector pBS TPp-A and -B recombination, protoplast transformation and flow cytometry measurement were carried out as described (Thévenin et al., 2012) (see supplementary material Fig. S3).

We acknowledge Professor Keiko Torii for kindly providing seeds of ice1-2 and Scrm-D; Professor Dominique Bergmann and Dr Peter Rogowsky for helpful comments; the ‘Plateforme de cytology et imagerie végétale’ at the Institute Jean-Pierre Bourgin for access to flow cytometry equipment; and Alexis Lacroix, Prisicilla Angelot and Yannick Rasmus for technical assistance.