Haploinsufficiency after successive loss of signaling reveals a role for ERECTA-family genes in Arabidopsis ovule development

The final size and shape of plant organs are determined by developmental programs that coordinate cell proliferation, cell expansion and cell-type differentiation (Mizukami,2001; Potter and Xu,2001). The Arabidopsis ovule has a relatively simple structure and a specific, well-defined differentiation pattern, making it a useful model for understanding the regulation of growth and organogenesis. The ovule originates from the placenta as an elongate protrusion with a defined proximal/distal axis that can be separated into three regions. The distal portion, called the nucellus, is the site of megasporogenesis and embryo sac development. The proximal region differentiates into the funiculus or stalk,which attaches the ovule to the carpel wall. Two integuments initiate at the central chalazal region and eventually envelop the nucellus and form the seed coat. Similar to the process of leaf laminar expansion, initiation and expansion of the integument requires the juxtaposition of abaxial and adaxial factors such as those encoded by PHABULOSA (PHB), KANADI1 (KAN1), KAN2 and KAN3, ABERRANT TESTA SHAPE (also known as KAN4 - TAIR), and the YABBY-family member INNER NO OUTER (Eshed et al.,2001; Eshed et al.,2004; McAbee et al.,2006; Sieber et al.,2004; Villanueva et al.,1999). In addition, transcription factors such as AINTEGUMENTA and NOZZLE (also known as SPOROCYTELESS), and the mitochondria ribosomal protein HUELLENLOS, have been identified as necessary for integument initiation(Elliott et al., 1996; Schneitz et al., 1998; Skinner et al., 2004; Villanueva et al., 1999).

A second group of loci affect the progression of integument growth after initiation. Several genes, including PRETTY FEW SEEDS 2(PFS2), SHORT INTEGUMENTS 1 and 2 (SIN1and 2) and TSO1, affect cell proliferation or expansion of the integuments, and their loss-of-function mutations result in reduction or loss of fertility (Hauser et al.,2000; Park et al.,2004; Park et al.,2005; Schneitz et al.,1997). Embryo sac failure is a secondary consequence of the absence of integuments; the gametophyte fails to develop in cases where the nucellus is not enclosed (Gasser et al.,1998). The cause of gametophyte abortion is not clear, but is likely to lie in the requirement for tight coordination of cell division and expansion within the ovule and communication between gametophytic and sporophytic tissue (Gasser et al.,1998). Relatively little is known about the genes involved in the coordination of growth or cell-cell communication within the ovule. This might be due to redundancy among the genes involved in this process, or to pleiotropic effects resulting from the loss of these genes. For instance, TSO1, TOUSLED (TSL) and SIN1 have multiple,sometimes detrimental, effects on vegetative and floral development in addition to ovule defects (Ehsan et al.,2004; Lang et al.,1994).

ERECTA (ER) and its two paralogs, ERECTA-LIKE 1(ERL1) and ERL2, regulate organ shape and inflorescence architecture and are members of the leucine-rich repeat receptor-like kinase(LRR-RLK) gene family in Arabidopsis(Shiu and Bleecker, 2001; Torii, 2004; Torii et al., 1996). Based on phylogeny, it was suggested that ERL1 and ERL2 evolved by recent gene duplication and are functionally related to ER, while maintaining overlapping but unique transcript expression patterns(Shpak et al., 2004). We report here that in the absence of functional ER and ERL1,Arabidopsis plants heterozygous at the ERL2 locus exhibit specific defects in integument development. Based on data from genetic and gene expression analysis, the ER family plays a key role in ovule development and fertility by regulating cell proliferation in the integuments. Our study highlights the unequal redundancy and dosage compensation among ER-family genes, and further reveals a potential molecular consequence of their reduced dosage for cell division control of integument development and fertility. Finally, genetic interactions of the ERfamily with PFS2, a WUSCHEL (WUS)-type homeodomain gene, revealed an unexpected intersection of two pathways required for proper integument growth and embryo sac development.

The Arabidopsis thaliana ecotype Columbia (Col) was used as wild type. All mutants are in the Col background unless otherwise indicated. The er-105, erl1-2 and erl2-1 mutations were described previously (Shpak et al.,2004). The erl1-1 allele (YJ133) was identified as an enhancer-trapped line by Dr John Bowman (University of California, Davis, CA)and kindly provided as a gift. The line was originally in the Landsberg erecta (Ler) background and outcrossed into Col three times. The erl1-4 allele was isolated from ethyl methanesulfonate-mutagenized Arabidopsis er-105 erl2 seeds and outcrossed to Col. The pfs2-1 allele was a gift from Dr Bernard Hauser (University of Florida, Gainesville, FL) and was in the Lerbackground and outcrossed to Col. The pfs2-1 mutant was crossed with the er-105 erl1-2 (+/-) erl2-1 mutant for further analysis. All plants were grown under long-day conditions as described previously(Shpak et al., 2003).

PCR-based genotyping for er-105, erl1-2 and erl2-1mutations was described previously (Shpak et al., 2004). Detection of the erl1-1 T-DNA insertion was performed with PCR primers: ERLK-185.rc,5′-CGTAGGTCTCCAATACGTGGA-3′ and GUS R,5′-CAGTTGCAACCACCTGTTGAT-3′. The erl1-4 mutation was detected by derived Cleaved Amplified Polymorphic sequences (dCAPs) using primers: erl1-4dcaps1151, 5′-GACGATGTTCACAACAGTGACTTGTGTTCTAG-3′and erl1-4dcaps1447.rc, 5′-CATCGAAATCAACAGAGAAAGAAAGGG-3′, and subsequent digestion with XbaI for 2 hours at 37°C, which cuts the mutant erl1-4 allele sequence.

Tissue preparation and in situ hybridization were performed as described previously (McAbee et al.,2006). Probe template construction was performed as follows. For the ERL1 probe template, the ERL1 sense and antisense probes were produced by linearizing plasmid pLJP562 (contains ERL1 kinase domain) with BamHI or XhoI and transcribing in vitro with T3 or T7 RNA polymerase, respectively. For the ER probe template, the ER sense and antisense probes were synthesized by linearizing pLJP131(contains the ER kinase domain and partial 3′UTR) with XhoI or BamHI and transcribing in vitro with T3 or T7 RNA polymerase, respectively. The WUS sense and antisense probes were produced by linearizing pSMB106 (nt 627-816 of the WUS cDNA) with EcoRI and XbaI and transcribing in vitro with T3 or T7 RNA polymerase, respectively. PFS2 probe template construction and hybridization conditions were according to Park et al.(Park et al., 2005). Sense and antisense probes for PHB were produced by linearizing plasmid pPHB (a gift from Dr Kiyotaka Okada, Kyoto University, Kyoto, Japan) with SacI or KpnI and transcribing in vitro with T7 or T3 RNA polymerase, respectively.

proERL1::GUS and proERL2::GUS constructs were described previously (Shpak et al.,2004). Histochemical staining for β-glucuronidase (GUS)activity was performed as described previously(Sessions et al., 1999).

Total RNA was isolated from Arabidopsis stage 12 gynoecia(Schneitz et al., 1995) or from ovules isolated from stage 12 gynoecia using the RNeasy Plant Mini Kit(Qiagen, Valencia, CA) and treated with DNase I (Amp Grade, Invitrogen). First-strand cDNAs were generated by ThermoScript Reverse Transcriptase(Invitrogen) and random hexamers using 1.0 μg of RNA according to manufacturer's instructions. Real-time PCR was performed with a Chromo4 thermocycler (BioRad, Hercules, CA) and analyzed using Opticon Monitor 3 software (BioRad). Triplicate reactions were performed in 25 μl total volume containing iTaq SYBR Green mix (BioRad) and 1.0 μl of the first-strand reaction. ACTIN 2 was used as an internal control in parallel reactions for each run. The relative level of expression of each target gene was calculated with respect to ACTIN 2 expression for each genotype. Real-time reactions were repeated in two biological replicates and the average fold-difference relative to wild type was calculated. Oligomers and DNA sequences used for PCR are shown in Table 1.

Pictures of plants were taken with an Olympus SZX12 dissecting microscope attached to an Olympus C-2020Z digital camera (Center Valley, PA). Tissue fixation, plastic embedding, sectioning, light- and scanning electron microscopy were performed as described previously(Shpak et al., 2003). To examine ovules using differential interference contrast (DIC) optics, carpels were dissected from flowers and incubated in 9:1 (v:v) ethanol:acetic acid overnight, rinsed with water and placed in choral hydrate (1:8:1 v:v glycerol:chloral hydrate:water). Cleared ovules and in situ hybridization sections were viewed under DIC optics using an Olympus BX51 microscope equipped with a DP70 digital camera.

Cell size was measured from images of plastic-embedded sections of wild-type and er erl1-2 erl2-1/+ carpels. The length and width of the first ten cells (nucellar to chalazal end) of the outer layer of the outer integument were measured using pixel distance in Photoshop Pro (Adobe Systems,San Jose, CA). Images used for these measurements corresponded to similar cross-sections through the nucellus of mature wild-type and er erl1-2 erl2-1/+ ovules. Cell numbers were counted from cleared ovules at stages 1-II, 2-IV, 3-II and 3-VI (Schneitz et al., 1995). At stages 2-IV, 3-II and 3-VI, counting started at the tip of the outer integument (nucellar end) and stopped where cells of the funiculus appeared obvious. At stage 1-II, all cells in the L1 layer of the ovule primordia were counted.

As reported previously, er, er erl1 and er erl2 null mutants developed compact but otherwise normal inflorescences, whereas the loss of all ER-family genes led to extremely dwarf and sterile plants with aberrant flowers disrupted in organ patterning and differentiation(Shpak et al., 2004)(Fig. 1A-D). To unravel the specific functions of ERL1 and ERL2 during inflorescence and floral organ growth, we investigated the phenotypes conferred by reduced dosages of ERL1 and ERL2 in the absence of ER. For this purpose, 100 self-fertilized progenies of er-105 erl1-2/+erl2-1/+, pre-selected for the presence of the erl1-2 T-DNA insertion, were grown to maturity. This population segregated seven extremely dwarf and sterile plants, consistent with the expected 8.3% ratio for the er-105 erl1-2 erl2-1 triple mutant (dwarf plants/total=7/100;χ ²=0.221, P=0.637). Seventeen additional plants from this population were completely sterile and produced no seeds. However, unlike the triple er-family mutants, they did not exhibit severe dwarfism,disrupted phyllotaxis or floral organ patterning defects. A reciprocal cross-pollination with wild type revealed that these plants were female sterile (data not shown).

To determine the genetic basis for this novel phenotype, we next analyzed the genotypes of all 100 plants in this population. All seventeen female-sterile, non-dwarf plants were homozygous for er-105 and erl1-2 and heterozygous for erl2-1 (i.e. er-105 erl1-2 erl2-1/+) (Fig. 1E). This was highly consistent with the expected ratio of 16.7% for this population(χ²=0.006, P=0.936). Unlike er-105 erl1-2 erl2-1/+, the er-105 erl1-2/+ erl2-1 plants were fully fertile(Fig. 1F). Reintroduction of functional copies of ER, ERL1 and ERL2 into respective loss-of-function mutants confirmed these data (see Table S1 in the supplementary material). The results indicate that, in the absence of ER and ERL1, ERL2 is haploinsufficient for female fertility,whereas ERL1 is haplosufficient in the absence of ER and ERL2. By contrast, the lack of defects in floral morphology (see Fig. S1 in the supplementary material) and phyllotaxis indicate that a single copy of ERL2 is sufficient for inflorescence elongation and floral patterning.

To further investigate the importance of ERL1 function in female fertility, we generated mutant combinations with two additional erl1alleles. The erl1-1 allele carries a T-DNA insertion 349 bases upstream of the initiation codon and represents a weak allele(Fig. 1I,J). The erl1-4 allele has a G-to-A substitution at nucleotide position +750. This creates a stop codon at amino acid 61, leading to a premature truncation of the protein prior to the LRR domain(Fig. 1I). Therefore, erl1-4 is a null allele.

The er-105 erl1-1 erl2-1, er-105 erl1-2 erl2-1 and er-105 erl1-4 erl2-1 triple-mutant plants were almost indistinguishable from each other (data not shown). Similar to er-105 erl1-2 er2-1/+ plants(Fig. 1E), er-105 erl1-4 erl2-1/+ plants were completely female sterile(Fig. 1G). By contrast, er-105 erl1-1 er2-1/+ plants were fertile and produced viable seed(Fig. 1H). These results demonstrate that a slight increase in ERL1 activity is sufficient to rescue the haploinsufficiency of ERL2. However, because erl1-1 failed to promote organ growth, proper floral patterning or fertility in the absence of ER and ERL2, the results highlight a strict ER-family signaling threshold required to promote female fertility.

To gain insight into the developmental basis of er-105 erl1-2 erl2-1/+ female sterility, we examined the morphology of mature ovules(stage 13 flowers, anthesis) (Smyth et al., 1990) (Fig. 2). Similar to wild type, er-105 erl1-2/+ erl2-1exhibited typical amphitropic ovule morphology(Fig. 2A,B,E). Ovules of er-105 erl1-1 erl2-1/+ were fertile, but had slightly shorter integuments than wild type, resulting in the micropyle resting slightly away from the funiculus (Fig. 2C). By contrast, the mature ovules of er-105 erl1-2 erl2-1/+ displayed aberrant morphology. Growth of both inner and outer integuments as well as the funiculus was dramatically compromised, leading to small, stunted ovules that often had an exposed nucellus (Fig. 2D,F). The embryo sac of mature er-105 erl1-2 erl2-1/+ovules was absent and replaced by a disorganized mass of small cells(Fig. 2F).

Initiation and polarity of integuments was not disrupted in er-105 erl1-2 erl2-1/+ (Fig. 2G). To determine whether integument initiation was defined normally, we used WUS and PHB as a nucellar regional marker and adaxial marker, respectively (Fig. 2H-K). WUS is expressed exclusively in the nucellus of wild-type ovules and promotes integument initiation in the chalazal region proximal to its expression domain(Groß-Hardt et al.,2002), whereas PHB expression marks the inner integument(Sieber et al., 2004). No alteration in WUS or PHB expression was detected between wild type and er-105 erl1-2 erl2-1/+. These data suggest that loss of ER-family signaling does not disrupt regional domain specification or radial patterning.

To determine the onset of developmental defects, wild-type and er-105 erl1-2 erl2-1/+ ovules were examined at sequential developmental stages(Fig. 3)(Schneitz et al., 1995). Early ovule initiation and patterning were normal in er-105 erl1-2 erl2-1/+, consistent with the normal PHB and WUSexpression patterns (Fig. 2I,K). Similar to wild type, ovule primordia initiated and elongated from the placental wall (stage 1-II; Fig. 3A,E,I). At stage 2-II to 2-V, outer and inner integument primordia arose from the chalaza and grew toward the nucellus (Fig. 3B,F,J). However, starting at stage 3-I, the cells of er-105 erl1-2 erl2-1/+ outer integuments had a more disorganized appearance than wild type (Fig. 3C,G,K). Enlargement of the megaspore mother cell (MMC) and tetrad formation proceeded as in wild type within the nucellus of er-105 erl1-2 erl2-1/+ ovules(Fig. 3I,M). However, mutant ovules arrested at a two-nucleate embryo sac(Fig. 3N). We never observed er-105 erl1-2 erl2-1/+ embryo sacs beyond this two-nucleate stage. Based on our observations, we conclude that progression of integument growth in the absence of ER is dependent on ERL1 and 2 in a dosage-dependent manner. At a minimum, ERL1 in a heterozygous state or ERL2 in a homozygous state is required.

Change in organ size reflects an alteration in cell size, cell number, or both. To understand the mechanism underlying the phenotype of er-105 erl1-2 erl2-1/+ ovules, we examined the size and number of cells in developing ovule integuments and compared them with wild type. In wild type,integument cells at the micropylar end were much more elongated than those at the chalazal end (Fig. 2E). The er erl1 erl2/+ ovules appeared to lack this characteristic polar elongation of integument cells (Fig. 2F). This made the outer integument cells of er-105 erl1-2 erl2-1/+ ovules less organized and slightly more uniform in length than wild type. However, no significant difference in outer integument cell size area was found between wild-type and mutant ovules(Fig. 4A-C).

We next assessed the possibility of endoreduplication in mutant integument cells by observing the nuclear size using 4′,6-diamidino-2-phenylindole(DAPI). Endoreduplication is the repetitive duplication of chromosomal DNA without mitosis or cytokinesis, resulting in an increase in cell ploidy level(Kondorosi et al., 2000). In plants, a positive correlation between ploidy level and cell size has been observed (Kondorosi et al.,2000; Melaragno et al.,1993). No qualitative difference in the size (area) of nuclei between wild-type and er-105 erl1-2 erl2-1/+ outer integument cells was observed (see Fig. S2 in the supplementary material), consistent with the absence of a cell size difference between wild-type and mutant integuments.

To investigate whether er-105 erl1-2 erl2-1/+ ovule integument defects were due to a decrease in cell division, we quantified the number of cells contributing to the epidermis at stage 1-II and to the outer integument at three subsequent stages (see Materials and methods). At stages 1-II and 2-II, no difference in cell number was observed between er-family combination mutants and wild type (Fig. 4D). However, er-105 erl1-2 erl2-1/+ outer integuments had significantly fewer cells per outer integument than wild type from stage 3-III (31% reduction) to maturity (57% reduction). A similar trend was observed for er-105 erl1-2/+ erl2-1 (15% reduction) and er erl1-1 erl2-1/+ (30% reduction) ovules, which had an intermediate number of integument cells between that of wild type and er-105 erl1-2 erl2-1/+at maturity (Fig. 4D). Together, these data provide evidence that a reduction in cell division is responsible for the abnormal integument morphology of er-105 erl1-2 erl2-1/+ ovules, and that the number of cell divisions is sensitive to the dosage of ERL1.

Because cell divisions were reduced in er-105 erl1-2 erl2-1/+ovules, we examined the expression of cell-cycle-regulatory genes to determine if their transcript accumulation is altered in er-105 erl1-2 erl2-1/+ovules. The cell cycle is regulated at multiple points by cyclin-dependent kinases (CDKs) that form complexes with appropriate cyclins(De Veylder et al., 2003; Dewitte and Murray, 2003). Plants have a large number of core cell-cycle genes including plant specific B-type CDKs that are regulated at the transcriptional level(De Veylder et al., 2003). Plant cyclin genes can be divided into four groups (A, B, D and H). The evolution of three A-type subclasses (A1, A2 and A3), compared with only one in animals, has led to the idea that A-type cyclins might fulfil plant-specific roles (Caubet-Gigot,2000; Dewitte and Murray,2003).

To detect alterations in expression levels of cell-cycle regulatory genes,we performed RT-PCR analysis on RNA from stage 12 carpels (see Fig. S3 in the supplementary material). According to Schneitz et al.(Schneitz et al., 1995), this stage of carpel corresponds to the point in ovule development when we observed noticeable changes in integument morphology between er-105 erl1-2 erl2-1/+ and wild type. Most of the B-type CDKs and cyclin A, B and D genes did not show a dramatic change in transcript abundance in ER-family loss-of-function mutants. However, CYCA2;2 showed a pronounced decrease in expression in er-105 erl1-2 erl2-1/+ (see Fig. S3 in the supplementary material; data not shown). We next quantified the relative decrease in CYCA2;2 transcripts using real-time PCR. We found that CYCA2;2 consistently shows a significant decrease in expression after loss of ER-family function in both carpels and individual ovules(Fig. 4E). Conversely, CYCA2;3 had the opposite trend, showing increased transcript accumulation after ER-family loss-of-function, perhaps owing to compensation (Fig. 4E). The results suggest that reduced levels of ER-family gene function may impact cell division through disruption of specific cell-cycle-regulatory gene expression. Alternatively, altered levels of these cyclin genes could be the consequence of reduced cell division produced by a yet unknown mechanism.

To determine the expression pattern of the ER family during ovule development, we first observed their promoter activity using reporterβ-glucuronidase (GUS) fusion constructs(Fig. 5)(Shpak et al., 2004). ERL1 and ERL2 promoter activity was first detected in the placental wall prior to ovule primordia initiation (data not shown). They became broadly active in ovules with a slight increase in activity in the central chalazal region (Fig. 5A-D). In mature ovules, faint GUS expression was restricted to the funiculus (data not shown). Interestingly, proERL1::GUS was notably stronger in the er-105 background than in wild type(Fig. 5E,F).

Because integument growth was sensitive to the dosage of ERL1 and because the promoter activity of ERL1 and ERL2 are similar,we focused our analysis of expression overlap on ER and ERL1. We analyzed the mRNA expression patterns of ER and ERL1 by in situ hybridization(Fig. 6). In reproductive structures, ERL1 transcripts were detected in floral meristems and primordia (see Fig. S4 in the supplementary material). During ovule development, ERL1 transcripts were detected throughout ovule primordia from initiation to stage 3-II, with increased hybridization intensity in the chalazal region and developing integuments(Fig. 6A-C). At maturity, ERL1 hybridization signal was still detectable in the integuments and funiculus (Fig. 6D).

Expression of ER was observed in floral meristems as well as the carpel wall and developing petals (see Fig. S4 in the supplementary material). Similar to ERL1, ER transcripts were detected broadly in the developing ovule, with more intense signal in the chalazal region and developing integuments (Fig. 6E-G). ER expression lessened later in development, but was still detectable in the carpel and ovule integuments at maturity(Fig. 6H). As the probes for ER and ERL1 did not cross-hybridize with each other (see Fig. S4 in the supplementary material), our results suggest that overlapping expression patterns of ER and ERL1 account for their redundant roles in ovule development.

Given the overlapping expression patterns of ER-family genes in the ovule and their ability to functionally substitute for one another(Shpak et al., 2004), we tested whether loss of specific ER-family gene function would lead to changes in expression of the other family members. To assess changes in transcript abundance, quantitative real-time RT-PCR analysis was performed on RNA from stage 12 carpels (Fig. 6Q). An increase in both ERL1 and ERL2transcripts was observed in the er-105 background as compared with wild type and, reciprocally, the accumulation of ER transcripts was higher in the erl1 erl2 mutant background. In situ hybridization determined that the spatial expression patterns of ER and ERL1 were not disrupted in the er-105 and erl1 erl2mutant backgrounds (Fig. 6I-P),respectively.

Taken together, the results suggest that transcriptional upregulation occurs in the absence of ER-family members, allowing each family member to functionally compensate for each other. This is consistent with the findings of Shpak et al. (Shpak et al.,2004) that ER-family members are able to substitute for each other if their temporal and spatial expression patterns overlap. Fertility in the null er single mutant or erl1 erl2 double mutant might require dosage compensation by the remaining family members to maintain signaling above a required threshold. This threshold is not met in the er-105 erl1 erl2/+ mutant, resulting in the arrest of integument growth.

PFS2, a WUS-related homeobox gene also known as WOX6, affects ovule patterning and regulates integument cell proliferation and gametophyte differentiation(Park et al., 2004; Park et al., 2005). It has been well-documented that the competing actions of the transcription factor WUS, and the CLAVATA (CLV) LRR-RLK signaling pathway, maintain meristem size in Arabidopsis (Brand et al.,2000). Numerous WOX genes display region-specific expression, but relatively little is known about their regulation(Haecker et al., 2004). To investigate whether a mechanism including PFS2 and ERsignaling directs integument size during ovule development, we produced the combination mutant er-105 erl1-2 erl2-1/+ pfs2-1 and characterized the phenotype.

The pfs2-1-null mutation causes the production of ovules with similar characteristics to er-105 erl1-2 erl2-1/+. pfs2-1 ovules have aberrant integument growth after initiation and embryo sac arrest(Park et al., 2004)(Fig. 7A). However, growth defects are variable in pfs2-1 mutants: some ovules complete various stages of embryo sac development and a low percentage of viable seed are produced (Park et al., 2005). Similar to pfs2-1 alone, we observed a variable degree of integument development in the er-105 erl1-2 erl2-1/+ pfs2-1 mutant(Fig. 7C,D). The phenotype ranged from an outer integument that covered the inner integument to a near absence of the outer integument. In addition, subtle changes in integument development were observed including lobing of the outer integument during development (Fig. 7C,D). Such variability in integument growth was not observed in er-105 erl1-2 erl2-1/+, in which the phenotype showed little variation(Fig. 7B). Surprisingly, some ovules of er-105 erl1-2 erl2-1/+ pfs2-1 developed beyond the two-celled nuclei stage typical of er-105 erl1-2 erl2-1/+ ovules,although no viable seed could be collected from these plants(Fig. 7E-G). Therefore, the loss of PFS2 function partially rescues the er-105 erl1-2 erl2-1/+ ovule defects.

Because pfs2-1 ovules show aberrant integument growth, we performed in situ hybridization using ER and ERL1 probes in the pfs2-1 ovules to determine whether loss of ER-family signaling contributed to the abnormal integument phenotype of pfs2-1 ovules(Fig. 7H-K). Loss or change in expression domain of ER or ERL1 was not detected in the pfs2-1 mutant, suggesting that the required threshold of ER-signaling is met in the pfs2-1 mutant.

The phenotypic rescue of er-105 erl1-2 erl2-1/+ ovules by additional pfs2 mutation implies that it is advantageous not to have PFS2 under the sub-threshold level of ER-family signaling. To test whether the er-105 erl1-2 erl2-1/+ phenotype was due in part to misexpression of PFS2, we performed in situ hybridization using a PFS2 probe (Fig. 7L-O). PFS2 expression was detected broadly in developing carpels and initiating ovule primordia (data not shown) with strong expression in the inner integument during initiation and growth(Fig. 7L-O). The expression domain of PFS2 was not altered in the er-105 erl1-2 erl2-1/+ovules (Fig. 7P,Q), indicating that the gametophytic defects in er-105 erl1-2 erl2-1/+ are not due to PFS2 misexpression. However, quantitative real-time RT-PCR detected an increase in PFS2 transcript abundance in er-105 erl1-2 erl2-1/+ stage 3-II ovules compared with wild type, indicating a potential inhibitory role for the ER family in PFS2 expression. Phenotypic rescue of er-105 erl1-2 erl2-1/+ by the addition of the pfs2 mutation suggests that ovule development requires a proper balance of both ER family and PFS2 function.

Gene duplication is a common occurrence and has led to the production of numerous gene families in Arabidopsis, one of the largest being the RLK-coding gene family, which has over 600 members(Shiu and Bleecker, 2001; Vision et al., 2000). Following a gene duplication event, decreased selection pressure on one of the gene copies can produce proteins with novel function or expression resulting in partial redundancy among gene family members(Hardtke et al., 2004). If the shared function of duplicated genes becomes unequally distributed between gene copies, unequal redundancy among the gene family members may result, as described for the ER family, where both erl1 and erl2 confer no growth phenotype but enhance the er growth defects in a unique manner (Kempin et al.,1995; Shpak et al.,2004). In many cases, unequal redundancy is produced by differences in the expression level or expression patterns between gene copies, where the duplicated gene(s) may have redundant function with the ancestral gene, but the expression might be lower or restricted. A good example is the three Arabidopsis brassinosteroid receptor genes. The major receptor, BRI1, is broadly expressed at high levels, whereas expression of BRL2 and BRL3 is restricted to vasculature(Cano-Delgado et al., 2004). Successive loss of family members enhances the phenotype of the ancestral gene and a true null phenotype is observed only in the absence of the entire family as has been shown for ER, ERL1 and ERL2(Shpak et al., 2004).

Our study uncovered unequal contributions and dosage compensation among gene family members with overlapping expression patterns. The most significant difference between er-105 erl1-2/+ erl2-1 and er-105 erl1-2 erl2-1/+ plants is the loss of female fertility owing to reduced cell division and outgrowth of the integuments in er-105 erl1-2 erl2-1/+. The expression pattern for all three family members overlaps in the distal,central and proximal regions, consistent with all three regions being reduced when successive family members are removed. An erl1-2 erl2-1 double or er single mutant shows no obvious defects in integument development, indicating that the upregulation of ER or ERL1and 2 in these backgrounds can maintain expression above the threshold necessary to direct integument development(Fig. 6Q). However, as successive family members are lost, low levels of ERL2 are insufficient to maintain signaling, and integument growth is compromised. Consistent with this dosage-dependence model, the ovules of er-105 erl1-2 erl2-1/+ are significantly larger and more developed than those of the ER-family triple mutant (see Fig. S5 in the supplementary material).

The ER-family LRR-RLKs probably maintain cell-cell communication, a reduction of which results in the irregular and reduced cell division during integument outgrowth in er-105 erl1-2 erl2-1/+ ovules. They do not regulate ovule primordia patterning or integument initiation, as these aspects are not affected even by the complete loss of ER-family function(Fig. 2 and see Fig. S5 in the supplementary material).

Our results suggest that the ER-family RLKs promote cell division after integument initiation through potential regulation of core cell-cycle genes,specifically CYCA2;2 (Fig. 4). Among over 30 cyclins present in Arabidopsis, ten belong to the A-type (Dewitte and Murray,2003; Vandepoele et al.,2002). Recent studies suggest that Arabidopsis cyclin A2 families act as negative regulators of endoreduplication(Burssens et al., 2000; Dewitte and Murray, 2003; Imai et al., 2006; Yu et al., 2003). For example,loss-of-function mutations in CYCA2;3 led to elevated ploidy levels in mature organs (Imai et al.,2006). Consistently, the dominant, activation-tagged mutation in INCREASED LEVEL OF POLYPLOIDY1, a conserved repressor of CYCA2 transcription, caused an increase in endoreduplication(Yoshizumi et al., 2006). However, cyclin A2 families may also function to regulate mitosis. In Medicago, CYCA2;2 (MedsaCYCA2;2) is required for meristem formation or activity but dispensable for endoreduplication(Roudier et al., 2000). Similar to Medsa;CYCA2;2, all four Arabidopsis cyclin A2 genes show high promoter activity in meristems and young organ primordia(Burssens et al., 2000; Imai et al., 2006). Therefore,they might redundantly control cell division in actively proliferating tissues, where the ER-family signaling pathway is required. Consistent with this hypothesis, flow-cytometric analysis of DNA content in er-105suggested possible cell-cycle arrest(Shpak et al., 2003).

All members of the ER family are expressed in overlapping domains in additional areas of the plant including shoot and inflorescence meristems and floral primordia (Shpak et al.,2004). The lack of any visual er-105 erl1-2 erl2-1/+phenotype in these tissues could be due to redundancy by other signaling factors that are not expressed in the ovule. Alternatively, ovule integuments might be particularly susceptible to cell-cell signaling disruption. Unlike meristems and developing leaves and floral organs, integuments are solely derived from the L1 layer (Schneitz et al., 1997). Therefore, cell proliferation in integuments is likely to occur with limited inter cell-layer communication. Such a scenario was suggested for loss-of-function mutants of the RLK gene ARABIDOPSIS CRINKLY4, which display localized defects in integument development,despite its expression in a wider variety of tissues(Gifford et al., 2003).

PFS2 is thought to control ovule primordial patterning by regulating the timing of cellular differentiation(Park et al., 2005). In pfs2-1, premature differentiation of the cells contributing to the gametophyte or integument primordia could account for decreased integument length and defects in the embryo sac (Park et al., 2004). The addition of pfs2-1 to er-105 erl1-2 erl2-1/+ mutations led to increased variability in both integument and gametophyte development, similar to pfs2-1 alone, indicating that the pfs2-1 mutation is epistatic to the er-family mutations. One explanation for these results is that the ER-family RLKs and PFS2 both contribute to integument outgrowth through distinct but interrelated mechanisms. The ER family are components of a signaling pathway that sustain organized cell divisions, potentially through the regulation of core cell-cycle genes such as CYCA2;2. Without activation of this signaling pathway, integument cells may differentiate before undergoing sufficient cell divisions. PFS2 might be responsible for maintaining integument cells in a state that allows them to respond to factors required for coordinated cell division. In this scenario, the loss of PFS2 function would be epistatic to the loss of the ER family.

WOX genes generally have a spatially restricted expression pattern(Matsumoto and Okada, 2001). Our in situ hybridization shows that PFS2 expression is spatially restricted to the inner integument (Fig. 7). Therefore, PFS2 might be required to properly coordinate the developmental states of the sporophytic integument tissues and gametophytic embryo sac. If so, the loss of PFS2 uncouples embryo sac development from abrogated integument growth under the sub-threshold level of ER-family signaling, therefore partially rescuing the embryo sac defects in erecta-105 erl1-2 erl2-1/+ ovules.

In the well-established model of CLV/WUS interaction in the shoot meristem, CLV3 limits the expression of WUS, while WUS in turn promotes the expression of CLV3, causing a feedback loop that maintains a stem cell population(Brand et al., 2000; Clark, 2001; Laux et al., 1996). Our results indicate that the ER family does not spatially restrict PFS2 expression and that factors yet to be identified are likely to be responsible for the limiting PFS2 expression. However, the ER family might restrict the expression levels of PFS2, and elevated expression of PFS2 might account for the abortion of gametophytes in er-105 erl1-2 erl2-1/+ ovules. Consistently, embryo sacs of er-105 erl1-2 erl2-1/+ are filled with a small group of cells, which resembles the phenotype of PFS2 overexpression described by Park et al.(Park et al., 2005).

Haploinsufficiency refers to a phenotype associated with the inactivation of a single allele leading to a half-normal amount of gene product, which is insufficient to maintain the wild-type phenotype. In this report, we have shown a novel phenotype for successive ER-family loss-of-function; ERL2 is haploinsufficient in the absence of ER and ERL1 to direct integument growth, but sufficient to drive floral organ growth, elongation and patterning. We further uncovered evidence of dosage compensation within the ER family and their potential regulatory relationships with the WOX-family transcription factor. Our work has identified the specific timing of ERL1 action during integument cell proliferation, which might provide a biological system to further elucidate the signaling components.

We thank Dr John Bowman for the erl1-1 enhancer-trap line, Dr Bernard Hauser for pfs2-1 seeds, Drs Kiyotaka Okada, Mitsuhiro Aida and Masao Tasaka for PHB probes, Dr Caroline Josefsson for assistance in RT-PCR analysis of erl1-1, Dr Kay Schneitz for advice on the design of the genetic screen and Dr Jessica Messmer-McAbee for commenting on the manuscript. This work was supported by the USDA CREES postdoctoral fellowship to L.J.P. and by the DOE grant (DE-FG02-03ER15448) to K.U.T. The DIC microscopy was supported by the CREST award from Japan Science and Technology Agency to K.U.T.