Roles of polarity determinants in ovule development


*For correspondence (fax +1 530 752 3085; e-mail


Ovules are the female reproductive structures that develop into seeds. Angiosperm ovules include one, or more commonly two, integuments that cover the nucellus and female gametophyte. Mutations in the Arabidopsis KANADI (KAN) and YABBY polarity genes result in amorphous or arrested integument growth, suggesting that polarity determinants play key roles in ovule development. We show that the class III homeodomain leucine zipper (HD-ZIPIII) genes CORONA (CNA), PHABULOSA (PHB) and PHAVOLUTA (PHV) are expressed adaxially in the inner integument during ovule development, independent of ABERRANT TESTA SHAPE (ATS, also known as KANADI4) activity. Loss of function of these genes leads to aberrant integument growth. Additionally, over-expression of PHB or PHV in ovules is not sufficient to repress ATS expression, and can produce phenotypes similar to those of the HD-ZIPIII loss-of-function lines. The absence of evidence of mutual negative regulation by KAN and HD-ZIPIII transcription factors is in contrast to known mechanisms in leaves. Loss of HD-ZIPIII activity can partially compensate for loss of ATS activity in the ats cna phb phv quadruple mutant, showing that CNA/PHB/PHV act in concert with ATS to control integument morphogenesis. In a parallel pathway, ATS acts with REVOLUTA (REV) to restrict expression of INNER NO OUTER (INO) and outer integument growth. Based on these expression and genetic studies, we propose a model in which a balance between the relative levels of adaxial/abaxial activities, rather than maintenance of boundaries of expression domains, is necessary to support laminar growth of the two integuments.


In flowering plants, ovules are critical female reproductive structures that develop into seeds. Angiosperm ovules include one, or more commonly two, integuments that cover the nucellus and female gametophyte. After fertilization, the integuments become the seed coat. Ovule ontogeny has been well characterized in Arabidopsis (Robinson-Beers et al., 1992; Schneitz et al., 1995), and many genes involved in ovule development have been identified (Skinner et al., 2004). Recent genetic studies have shown that mutations in putative polarity determinants, such as members of the YABBY and KANADI gene families, result in amorphous or arrested integument growth (Villanueva et al., 1999; Eshed et al., 2001; McAbee et al., 2006), suggesting that such determinants play key roles in ovule development.

ABERRANT TESTA SHAPE (ATS, also referred to as KANADI4, KAN4) is a member of the KANADI gene family that is necessary for laminar extension of the inner integument and for maintaining integument separation (McAbee et al., 2006). In ats mutant ovules, the two integuments fail to originate as separate structures, resulting in a single ‘fused’ integument, and leading to aberrant heart-shaped seeds (Léon-Kloosterziel et al., 1994; McAbee et al., 2006). Loss of function of two other KANADI family members, KAN1 and KAN2, results in an amorphous outer integument and a normal inner integument, implying that these polarity determinants are necessary for laminar extension of the outer integument (Eshed et al., 2004; McAbee et al., 2006). A dominant allele of one class III homeodomain leucine zipper (HD-ZIPIII) gene, PHABULOSA (PHB), displays incomplete integument formation, implying that these transcription factors may also play important roles in ovule patterning and growth (McConnell and Barton, 1998).

Based on current knowledge regarding the expression and roles of polarity genes during leaf morphogenesis, a model has been proposed to explain initiation and extension of integuments and phenotypes of ovule mutants such as ats (McAbee et al., 2006). In leaves, combined expression of HD-ZIPIII transcription factors provides adaxial identity, while YABBY and KANADI family members are abaxial determinants (McConnell and Barton, 1998; Siegfried et al., 1999; Kerstetter et al., 2001; McConnell et al., 2001; Bowman et al., 2002; Emery et al., 2003). Loss of polarity function can result in arrested and/or radialized lateral organs, leading to the hypothesis that juxtaposition of adaxial and abaxial characters is required in order to properly define locations of laminar growth (Waites and Hudson, 1995; Waites et al., 1998; Eshed et al., 2001, 2004; Bowman et al., 2002). McAbee et al. (2006) proposed that ATS acts as an abaxial determinant in the inner integument, while KAN1 and KAN2 act in conjunction with a YABBY gene (INNER NO OUTER, INO) to specify abaxial cell fate in the outer integument. Furthermore, McAbee et al. (2006) also hypothesized that HD-ZIPIII transcription factors act in both the outer and inner integuments as adaxial determinants.

While there is evidence to support the idea that abaxial determinants are required for initiation and maintenance of integument growth (Villanueva et al., 1999; Eshed et al., 2001; McAbee et al., 2006), a role for adaxial determinants in ovules has not yet been demonstrated. We sought to further define the role of polarity establishment in integument growth through expression analyses and genetic and transgenic studies of HD-ZIPIII and ATS transcription factors.

In addition to their roles in leaves, we have found that CORONA (CNA), PHB, PHAVOLUTA (PHV) and REVOLUTA (REV) also regulate ovule development. We show that these HD-ZIPIII transcription factors act in concert with ATS to control patterning and laminar growth of both the inner and outer integument. These results provide evidence that a polarity establishment pathway is required for integument formation, and also reveal differences between the pathways utilized in ovule and leaf development. In light of these new findings, we propose a model in which a balance between levels of polarity determinants acts to mediate integument development.


The HD-ZIPIII genes PHB, PHV and CNA are expressed in a polar fashion during ovule development

Based on previous genetic studies (Eshed et al., 2001; McAbee et al., 2006), we hypothesized that HD-ZIPIII genes may function as adaxial determinants during integument morphogenesis. There are five HD-ZIPIII genes in Arabidopsis: CNA/ATHB15, PHB, PHV, REV and ATHB8. The patterns of expression ofPHB and REV have previously been examined during ovule development (Sieber et al., 2004a). Here we focus on expression of PHV and CNA, and re-examine PHB as a control, because of their demonstrated roles in leaf development and possible role in ovule development (McConnell and Barton, 1998; Sieber et al., 2004a). ATHB8 expression has previously been shown to be primarily associated with vascular development (Prigge et al., 2005), and this gene was therefore not examined in this study. Expression patterns of PHB, PHV and CNA were examined in wild-type ovules using in situ hybridization with gene-specific probes. In stage 2 ovules (stages according to Schneitz et al., 1995), PHB mRNA was detected only in the inner integument, specifically in the cell layer adjacent to the nucellus (Figure 1a and Figure S1a,b). In later stages, hybridization was also seen in the vasculature of wild-type ovules (data not shown). This expression pattern is consistent with a previous report (Sieber et al., 2004a). PHV and CNA were expressed in similar patterns in ovules (Figure 1c,e, respectively, and Figure S1e, f, i), with inner integument-specific expression being present as early as stage 1 ovules (Figure 1e). Together, the PHB, PHV and CNA in situ hybridizations indicate an adaxial pattern of expression in the inner integument (as predicted by McAbee et al., 2006).

Figure 1.

 The polar expression patterns of HD-ZIPIII genes are independent of ATS activity.
(a, c, e) Wild-type Ler ovules; (b, d, f) ats-1 mutant ovules; (a, b) in situ hybridization with antisense PHB probe; (c, d) in situ hybridization with antisense PHV probe; (e, f) in situ hybridization with antisense CNA probe. Abbreviations: n, nucellus; ii, inner integument; oi, outer integument; i, integument.

During polarity establishment of leaves, HD-ZIPIII expression patterns are refined through a combination of microRNA regulation and repression by KANADI genes (Bowman et al., 2002; Kidner and Timmermans, 2007). Given that ATS is the only KANADI gene that is known to be active in the inner integument, we wished to test whether ATS is required to restrict the expression of PHB, PHV and/or CNA during ovule development. In loss-of-function ats ovules, we observed expression patterns for PHB, PHV and CNA that did not differ from those observed in wild-type (for PHB, compare Figure 1b with Figure 1a; for PHV, compare Figure 1d with Figure 1c; for CNA, compare Figure 1f with Figure 1e; see Figure S1 for additional comparisons). This suggests that ATS activity is not required to confinedPHB, PHV or CNA expression patterns during ovule development.

PHB, PHV and CNA are required for integument morphogenesis

In order to examine the collective role(s) of PHB, PHV and CNA during ovule development, we used a genetic approach. These genes overlap in expression patterns as well as function, and therefore single and double mutants appear phenotypically wild-type (Prigge et al., 2005). However, a recent study of these genes described the cna-2 phb-13 phv-11 triple mutant as having ‘short integuments’ in addition to other developmental defects such as extra carpels (Prigge et al., 2005).

In our examination of cna-2 phb-13 phv-11, we found the ovule phenotype to be highly variable (Figure 2). In some cases, cna-2 phb-13 phv-11 ovules initiated both integuments normally, resulting in nearly wild-type ovules (Figure 2a,e, bottom right corner). More frequently, ovules exhibited a protruding inner integument and a superficially shorter outer integument at maturity (Figure 2b,f, top right corner). This indicates a disruption in the coordination of timing of initiation and/or growth of the two integuments. Despite this morphogenetic disconnect between the two integuments, the outer integument still retained its anatropous form. We also observed ovules that phenocopied the ovules of ino mutants (Baker et al., 1997; Villanueva et al., 1999) (Figure 2c e), with an inner integument and an absent or rudimentary outer integument. Ovules with an amorphous, globular structure in place of the integuments and an exposed nucellus were also commonly observed (Figure 2d–f). In general, it appears that the combined loss of CNA, PHB and PHV function leads to abnormal integument development. Surprisingly, this effect is not restricted to the inner integument, as would be predicted from the ovule expression patterns shown in Figure 1.

Figure 2.

 HD-ZIPIII genes are required for integument morphogenesis.
(a–f) cna phb phv triple mutant ovules; (g–i) phv-1d gain-of-function ovules. (a–d) Whole-mount ovule clearings; (e–i) scanning electron micrographs. Scale bar = 20 μm (a–d, i), 50 μm (e, f, h) or 5 μm (g). Abbreviations: n, nucellus; oi, outer integument; ii, inner integument; f, funiculus; ai, amorphous integument; i, integument.

Gain of function phb-1d mutants display ectopic PHB expression as a result of microRNA insensitivity (Reinhart et al., 2002; Rhoades et al., 2002). In phb-1d ovules, the timing of integument initiation and growth is disrupted, producing ovules with elongated inner integuments (McConnell and Barton, 1998; McConnell et al., 2001; Sieber et al., 2004a). We examined phv-1d ovules in comparison with those of phb-1d. In phv-1d plants, we observed three classes of ovules: phb-1d-like, ino-like and wild-type (Figure 2h,i). Because these dominant mutations represent microRNA-resistant alleles of PHB and PHV (McConnell et al., 2001), these data imply that proper regulation of PHB and PHV expression patterns via miR165/166 action is required for normal integument development. Additionally, these phenotypes were similar to those observed in the cna phb phv triple loss-of-function mutant, suggesting that a deviation from the normal level of adaxial activity produced by PHB, PHV and CNA may directly influence integument morphogenesis.

ATS and HD-ZIPIII functions are required to maintain integument development

We crossed cna-2 phb-13 phv-11 with ats-3 phb-6 phv-5 rev-9/+ in order to further examine the roles of HD-ZIPIII genes in the context of ATS function. We were able to identify ats cna phb phv and ats cna phb phv rev/+ mutants in the F2 population by PCR-based genotyping. Less than 10% of ats cna phb phv ovules had a wild-type appearance (Figure 3d), and no wild-type ovules were observed in ats cna phb phv rev/+. More frequently, ats cna phb phv ovules exhibited arrested outer integument growth (Figure 3e,f) and partial inner integument growth (Figure 3e,f; note the naked nucellus in Figure 3f and the amorphous inner integument in Figure 3e). Loss of one copy of REV in this mutant background enhances these defects, producing ovules with arrested inner and outer integument growth (Figure 3g,h). While both mutant combinations displayed a range of phenotypes, the ats cna phb phv rev/+ mutants were more severely affected than the ats cna phb phv mutants (compare Figure 3g,h and d–f), suggesting that REV activity does contribute to integument growth.

Figure 3.

 Loss of function of CNA, PHB and PHV can suppress the Ats phenotype.
(a, b) Whole-mount clearings of ats; (c–f) whole-mount clearings of ats cna phb phv ovules; (g, h) whole-mount clearings of ats cna phb phv rev/+ ovules. Scale bar = 5 μm (a, c) or 10 μm (b, d–h). Abbreviations: n, nucellus; oi, outer integument; ii, inner integument; f, funiculus.

Notably, combined loss of CNA, PHB and PHV function can suppress the integument fusion Ats phenotype. In ats ovules, the integuments are congenitally fused, resulting in a unitegmic ovule (McAbee et al., 2006) (Figure 3a). However, ats cna phb phv and ats cna phb phv rev/+ ovules initiated separate inner and outer integuments (Figure 3c and e–h, respectively; compare Figure 3a and c). It is worthwhile noting that this suppression is not observed in other mutant combinations tested, i.e. ats phb phv rev/+ plants (see Figure 4), suggesting that loss of CNA activity (in the absence of PHB and PHV) may be a predominant component of the suppression phenotype. Together, these mutant analyses provide genetic evidence that CNA, PHB, PHV and REV together with ATS are required to sustain normal integument growth.

Figure 4.

 Genetic interactions between HD-ZIPIII genes and ATS.
(a, b) phb phv rev/+; (c, d) ats single mutant; (e, f) ats phv double mutant; (g, h) ats phb double mutant; (i, j) ats phb phv triple mutant; (k, l) ats phb phv rev/+; (m) wild-type Ler; (n) phb phv rev/+; (o) ats phb phv rev/+. (a–l) Whole-mount clearings of ovules; (m–o) in situ hybridizations with antisense INO probe. Scale bar = 5 μm (a, c, e, g, i, k, m–o) or 10 μm (b, d, f, h, j, l). The black arrowhead in (m)–(o) indicates the gynoapical side of the ovule. Abbreviations: n, nucellus; oi, outer integument; ii, inner integument; f, funiculus; i, integument.

ATS and HD-ZIPIII genes interact to negatively regulate INO expression

Of the five HD-ZIPIII genes, only REV has been shown to be expressed in both integuments (Sieber et al., 2004b). Given this difference in expression domains, we wisheed to examine the roles of REV in the absence of ATS. We crossed phb-6 phv-5 rev-9/+ plants, which have wild-type ovules (Figure 4a,b), with ats-3 plants (Figure 4c,d) in order to examine whether or not these genes act in the same genetic pathway during integument development. In segregating F2 progeny from an ats-3/+ phb-6/+ phv-5/+ rev-9/+ parent, we were able to evaluate the phenotypes of desired mutant combinations after PCR-based genotyping. These mutant studies suggest that PHB, PHV and REV may have overlapping activities with ATS, as ats phv (Figure 4e, f), ats phb (Figure 4g, h) and ats phb phv (Figure 4i, j) mutants all show an intermediate degree of symmetrical integument growth (compare Figure 4f, h, j with Figure 4l). Additionally, ats phb phv rev/+ ovules are more severely affected than ats phb phv ovules (compare Figure 4k with Figure 4i, and Figure 4l with Figure 4j), and exhibit completely symmetrical outer integument development (Figure 4h,l). These mutant combinations also have significantly reduced seed set compared to wild-type (compare Figure S2d with Figure S2e,f). Furthermore, suppression of ats does not occur in the ats phb phv rev/+ mutant, in contrast to the ats phb phv cna quadruple mutant (Figure 3). This phenotypic difference suggests that both CNA and REV may act in concert with PHB and PHV in different ways during ovule development. While this hypothesis is consistent with previously demonstrated differences in function between CNA and REV (Prigge et al., 2005), it is important to note that it is based on our ability to examine only partial loss of REV function in the absence of PHB and PHV because true phb phv rev triple mutants are seedling-lethal (Emery et al., 2003).

The ats phb phv rev/+ ovule morphology is reminiscent of that of superman (sup) ovules, in which symmetrical growth results from ectopic INO expression on the gynoapical side of the developing ovule (Meister et al., 2002). We examined expression of INO in wild-type, phb phv rev/+ and ats phb phv rev/+ ovules to determine whether similar expression could account for the observed phenotype. As in wild-type, INO was expressed only on the gynobasal side of phb phv rev/+ ovules (Figure 4n compared to Figure 4m). In contrast, ats phb phv rev/+ ovules showed ectopic INO expression on the gynoapical side of the ovule (Figure 4o, arrow) in addition to the normal gynobasal location. The expression pattern of INO in ats ovules did not differ from that in wild-type ovules (McAbee et al., 2006), so the observed difference in expression patterns represents a synergistic effect between ATS and REV, PHB and PHV.

ATS expression is unaltered in phb-1d ovules

Antagonism between KAN and HD-ZIPIII genes occurs during leaf development to delineate abaxial–adaxial boundaries (Bowman et al., 2002; Kidner and Timmermans, 2007). In gain-of-function alleles of HD-ZIPIII genes such as phb-1d and phv-1d, the leaf tissue becomes adaxialized via repression of abaxial factors, and KAN expression is reduced (Eshed et al., 2001, 2004; Kerstetter et al., 2001; Emery et al., 2003). We examined ATS expression in phb-1d ovules using in situ hybridization to determine whether a similar mechanism may be acting in ovules. In wild-type ovules, ATS expression is first seen in boundary cells between the inner and outer integument (Figure 5a), and later becomes restricted to the inner integument (McAbee et al., 2006). ATS is expressed normally in phb-1d and phv-1d ovules (Figure 5b, c, and data not shown), implying that these HD-ZIPIII transcription factors at least are not sufficient to negatively regulate ATS when ectopically expressed during integument development. This suggests that the canonical repressive interactions between KAN and HD-ZIPIII genes in lateral organs may not be reiterated in ovules, at least with respect to ATS and PHB or PHV.

Figure 5.

ATS is expressed normally in phb-1d mutant ovules.
(a, d) Wild-type Ler ovules; (b, c, e) phb1d ovules. (a–c) In situ hybridizations with antisense ATS probe; (d, e) scanning electron micrographs. Scale bars = 5 μm. Abbreviations: n, nucellus; oi, outer integument; ii, inner integument.

Ectopic expression of ATS can arrest growth of the outer integument

Mis-expression of KAN genes can lead to arrest of organ growth (Kerstetter et al., 2001; Emery et al., 2003). We used the LHG4>>OP system (Liu and Meinke, 1998; Moore et al., 1998) to ectopically express ATS across the chalaza during ovule development under the control of the ANT promoter, which is active in the region from which both integuments form (Elliott et al., 1996). A line harboring a PANT:LHG4 transgene (Gross-Hardt et al., 2002) was crossed into POP:ATS, and the resulting ANT>>ATS F1 plants were evaluated. Two phenotypic classes of ovules were observed in these F1 plants: wild-type (Figure 6b) and ino-like (Figure 6c, d). The ino-like ovules had an inner integument and a reduced/absent outer integument (Figure 6c, d). The lack of complete penetrance of the ino-like phenotype may be due to weak expression from either the ANT:LHG4 and/or OP:ATS transgenes. These data suggest that expression of ATS in the chalaza can lead to outer integument arrest, possibly through negative regulation of INO, which is consistent with our mutant studies (Figure 4).

Figure 6.

 Ectopic expression of ATS can arrest growth of the outer integument.
Whole-mount clearings of ANT>>ATS ovules. (a) Integument initiation. (b–d) Mature ovule phenotypes: (b) wild-type ovule; (c,d) ino-like ovules. Abbreviations: i, integuments; ii, inner integument; oi, outer integument; mp, micropyle; f, funiculus.


Adaxial expression of HD-ZIPIII genes in the inner integument

While roles for KAN (ATS, KAN1 and KAN2) and YABBY (INO) genes in ovule development have been described, the functions of adaxial determinants in this process are less well understood. In leaves, the HD-ZIPIII expression domains are adjacent to the KAN expression domains, and this arrangement promotes laminar growth (Eshed et al., 2001; Kidner and Timmermans, 2007). Using in situ hybridization, we show that the HD-ZIPIII genes PHB, PHV and CNA are expressed in a polar fashion in the inner integument, with their mRNA accumulating in the cell layer adjacent to the nucellus (which later differentiates into the endothelium). Thus PHB, PHV and CNA expression was juxtaposed with ATS expression in ovules (Figure 1), consistent with the polarity model proposed by McAbee et al. (2006). In contrast to the other three examined HD-ZIPIII genes, REV is expressed broadly in both integuments (Sieber et al., 2004b), and thus does not conform to the proposed model.

HD-ZIPIII expression expands abaxially in leaves of kan mutants, demonstrating that KAN genes negatively regulate HD-ZIPIII expression and contribute to confinement of HD-ZIPIII expression to adaxial domains (Eshed et al., 2001, 2004; Kerstetter et al., 2001; Bowman et al., 2002). However, in ovules, it appears that PHB, PHV and CNA expression patterns are not governed by ATS activity because they are unaltered in ats ovules (Figure 1).

McAbee et al. (2006) hypothesized that the absence of abaxial function in ats mutants creates a single adaxial/abaxial boundary in the ovule, rather than two separate boundaries, leading to formation of a single integument. They further hypothesized that loss of ATS activity could lead to expansion of the ‘adaxial’ factor(s) expression domain. Our observation that PHB, PHV and CNA expression patterns are unchanged in ats ovules (Figure 1) conflicts with this hypothesis. Rather, it appears that loss of the ATS abaxial boundary function appears to be sufficient to produce the observed integument fusion.

As ATS is not responsible for determining the pattern ofPHB, PHV or CNA expression in ovules, what factor(s) could contribute to the difference in their expression patterns from that of REV in this structure? In leaves, patterning of HD-ZIPIII mRNA accumulation occurs in part through negative regulation by miR165/166 (Kidner and Timmermans, 2007). In ovules, the PHB promoter appears to be active in both integuments based on GUS activity in the phb-6 enhancer trap line (Figure S3). Because this promoter activity does not match the mRNA distribution pattern (Figure 1 and Sieber et al., 2004b), miR165/166 could contribute to restriction of PHB expression in ovules. In addition, the miRNA-resistant phb-1d mutant displays an expanded expression domain relative to the wild-type (Sieber et al., 2004b), further implicating miRNA in regulation of this gene. Although all five HD-ZIPIII genes share the miR165/166 recognition sequence, miR166g has been shown to have differential affects on HD-ZIPIII transcripts (Williams et al., 2005). Over-expression of miR166g in the jabba-1d mutant led to down-regulation of PHB, PHV and CNA mRNAs, while REV expression was increased (Williams et al., 2005). Based on these data, one hypothesis that could account for the differences in PHB, PHV and CNA mRNA distributions compared to the REV expression domain could be differential sensitivity to miR165/166 action.

The LITTLE ZIPPER (ZPR) proteins, a novel family of leucine zipper-containing proteins, have recently been proposed to negatively influence HD-ZIPIII activity and expression (Wenkel et al., 2007; Kim et al., 2008). The roles of ZPR genes in ovules have not yet been evaluated, but differential activity of ZPR proteins on the HD-ZIPIII genes represents another hypothesis for the different expression domains of REV and PHB/PHV/CNA.

HD-ZIPIII genes are required for patterning and growth during ovule development

Loss of PHB, PHV and CNA led to abnormal ovule development characterized by arrested or amorphous inner and outer integuments. It is curious that both integuments were affected when expression of these genes was only detected in the inner integument (Figure 1 and Sieber et al., 2004b). There are several possible explanations for this combination of observations. It is possible that in situ hybridization (Figure 1 and Sieber et al., 2004b) is insufficiently sensitive to detect low levels of PHB, PHV and/or CNA mRNA that may be present in the outer integument. Another possibility derives from the order and timing of integument formation. Inner integument initiation precedes initiation of the outer integument, and development of both structures is coordinated (Schneitz, 1999). If inner integument patterning is unbalanced, this could have an impact on the quality of outer integument growth by a domino effect. This type of non-cell autonomous action has been described for other transcription factors that are active in ovules, such as WUSCHEL (Gross-Hardt et al., 2002). An additional influencing factor could be the production/perception of hormones such as auxin. Recent studies on ARF6 and ARF8 indicate that auxin perception and responsiveness contribute to integument formation (Wu et al., 2006). Given that auxin signaling during embryogenesis appear to be mediated by KAN and HD-ZIPIII activity via PIN1 localization (Izhaki and Bowman, 2007), the same could be true in ovules, and alterations in inner integument development could alter the hormone environment in ways that would affect initiation and growth of the outer integument.

Reduction in outer integument growth was also observed in both phb-1d (McConnell and Barton, 1998) and phv-1d (Figure 2g–i), and inner integument defects were observed in phv-1d (Figure 2g–i). Over-production (ectopic production) of the products of these two genes could therefore produce effects similar to those resulting from a decrease in HD-ZIPIII function (e. g. Figure 2b). This seemingly paradoxical observation may be explained if an appropriate balance between adaxial- and abaxial-promoting activities is necessary for proper integument growth, and imbalance in either direction results in disruption of this process.

Loss of CNA/PHB/PHV can suppress both aspects of the Ats- phenotype in the ats mutant background (Figure 3). The observation that simultaneous loss of abaxial and adaxial functions restores inner integument growth and integument separation (Figure 3) indicates that there may be additional adaxial/abaxial factors that are active in ovules, or that other functions can substitute for the abaxial/abaxial juxtaposition when both classes of factors are absent. Indeed, loss of REV activity in this background reduces inner integument growth, suggesting that REV may be one of the factors that can compensate for loss of CNA, PHB and PHV. That loss of adaxial activity can mitigate the effects of loss of abaxial activity is consistent with our hypothesis that an appropriate balance between the levels of these two activities is critical for normal laminar extension of the integuments.

We also observed that ectopic expression of ATS in the chalaza during ovule development can lead to arrest of outer integument growth (Figure 6). These data are also consistent with the concept that an appropriate balance between levels of KAN and HD-ZIPIII functions must be maintained in order to promote integument growth, but other mechanisms are also possible. For example, ectopic KAN and ATS expression can lead to meristem arrest (Emery et al., 2003; Kerstetter et al., 2001 and unpublished data), and growth arrest may be a general activity of KAN proteins.

A balance model for the adaxial/abaxial determinants underlying integument morphogenesis

Based on these novel expression and genetic data, we have refined our model for integument growth (Figure 7) from that published previously (McAbee et al., 2006). We had previously proposed that ATS acts in juxtaposition to a hypothetical adaxial function to promote inner integument growth. We can now say with confidence that this adaxial function includes the activities of the HD-ZIPIII family members PHB, PHV and CNA, which are expressed adaxially in the inner integument (Figure 7a). These expression patterns may be established through a combination of promoter activities, miR165/166 expression and/or ZPR action, but are independent of ATS activity. The adaxial activity also includes REV function, even though this gene is expressed more broadly across the chalaza (Sieber et al., 2004b) (Figure 7a). Support for such a role for REV comes from comparison of ats cna phb phv rev/+ ovules and ats cna phb phv ovules (Figure 3). The loss of one copy of REV enhances the ats cna phb phv phenotype by negatively impacting inner integument growth. ATS is shown to be a critical component of the abaxial function, but we also hypothesize the existence of an additional abaxial function because inner integument growth and integument separation are restored by certain mutant combinations with ats. These results can be explained by an abaxial/adaxial juxtaposition model if the model is modified to hypothesize that integument growth depends on a proper balance between abaxial and adaxial activities, rather than the absolute levels of activities (Figure 7b). The observation that ATS expression remains unchanged in the absence of HD-ZIPIII, and vice versa, shows that, in contrast to the situation in leaves, the abaxial and adaxial functions are not mutually suppressive.

Figure 7.

 Balance model for roles of polarity determinants in ovule development.
(a) Expression summary of polarity determinants required for integument growth. During early ovule development, CNA, PHB and PHV (red), REV (all colored regions), ATS (blue), ‘?’ unknown abaxial factor(s) that may share expression domains with ATS, and INO (green) mRNAs show specific patterns of expression along the chalaza. KAN1 and KAN2 (green) are predicted to be expressed in a similar fashion to INO. Once these patterns are established, persistent expression/action of all these transcription factors is required for proper integument growth.
(b) The proposed polarity balance model. In the inner integument, ATS and the proposed additional abaxial factor (‘?’) act in balanced opposition to CNA/PHB/PHV and REV to promote inner integument growth. Disruption of this balance by a decrease or increase in either function alters the growth plane, leading to aberrant integument growth. Initiation of the inner integument precedes outer integument initiation in Arabidopsis, and thus growth of the outer integument is directly influenced by inner integument morphogenesis. In the outer integument, the abaxial activities of INO, KAN1 and KAN2 act in balance with REV to control morphogenesis.

The earlier model (McAbee et al., 2006) and the model for leaf development (Bowman et al., 2002; Kidner and Timmermans, 2007) predict a progressive loss of integument growth with progressive loss of abaxial/adaxial factors. However, our observation that loss of HD-ZIPIII activity can partially compensate for loss of ATS activity is inconsistent with these models. The revised balance model can explain these results and provide an explanation of how similar phenotypes can result from loss of function and ectopic expression of HD-ZIPIII genes. Whether such a balance mechanism also acts in leaves, or whether it represents a further difference between leaves and integuments has yet to be determined. Another aspect of the balance model relates to the coordinated growth of the inner and outer integuments. Loss of apparently inner integument-specific gene functions leads to disruption of outer integument growth (Figures 2 and 3, and McAbee et al., 2006), suggesting that inner integument growth positively contributes to outer integument growth. A final aspect of the model is that the pattern and growth are self-reinforcing. Once an initial state of appropriately balanced adaxial/abaxial definition has been achieved, it will be subsequently maintained. This could explain the range of phenotypes observed among ovules of individual polarity mutants, where small stochastic variations in the initial levels of determinants could result in different final outcomes.

Patterning roles of REV and ATS during integument morphogenesis

While a more complete understanding of REV function in ovules is lacking due to the inability to examine phb phv rev ovules, we have found that, in the absence of ATS, phb phv rev/+ ovules become sup-like (Figure 4). We attribute this phenotype to ectopic INO expression (Figure 4). Because we did not observe this same phenotype in ats phb phv cna ovules, and REV is expressed differently from its paralogs in ovules, we propose that REV may have a unique function in conjunction with ATS in patterning INO expression during integument initiation. This mechanism may be connected with SPOROCYTELESS/NOZZLE action, which was previously shown to act with ATS to regulate INO (Balasubramanian and Schneitz, 2002). Although KAN1 and KAN2 (the only other KANADI genes that appear to play a role in ovule development) also participate in the control of outer integument formation, expression patterns of these genes in ovules have not been detectable to date (D.R.K. and C.S.G., unpublished results).

How leaf-like are integuments?

The fossil record indicates that the origin of ovules was contemporaneous with the origin of leaves (Andrews, 1963; Herr, 1995; Gasser et al., 1998). According to the telome theory, the inner integument is homologous to lateral sterile or sterilized structures (borne on a reproductive telome truss, with the nucellus being homologous to the apical fertile telome). The outer integument was gained later in plant history, on the stem lineage leading to angiosperms, possibly by transformation from a leaf-like structure known as a cupule (Gasser et al., 1998; Doyle, 2006).

Separate origins for both integuments are also supported at the molecular level, as the control of inner and outer integument development occurs through different genes (Skinner et al., 2004; McAbee et al., 2006 and this study). For instance, ATS and CNA/PHB/PHV drive inner integument growth, while INO and KAN1/2 regulate outer integument growth, and REV is an apparent additional participant in both cases (Figure 7). Thus the inner integument shares process homology with leaves, in that it is not directly derived from a leaf but it does utilize the same set of gene families (KANADI and HD-ZIPIII) to control development. On the other hand, the outer integument is believed to share structural homology with leaves (being most likely derived from a cupule; Doyle, 2006). The outer integument developmental genetic program would therefore be expected to also include a YABBY gene, in this case the diverged family member INO. If REV acts as the corresponding adaxial factor in the outer integument, it may do so independently of a delineated adaxial/abaxial expression boundary (Sieber et al., 2004b). Current phylogenetic analyses of these three gene families (KAN, HD-ZIPIII and YABBY) are consistent with these hypotheses, with the origin of the KAN and HD-ZIPIII lineages pre-dating seeds and leaves, and YABBY genes originating with seed plants (Floyd and Bowman, 2007).

Together, our studies and those of previous researchers (Balasubramanian and Schneitz, 2002; Sieber et al., 2004b; McAbee et al., 2006) demonstrate that, while there are similarities between integument and leaf development, there are also marked differences. Although these organs have distinct evolutionary origins, a common set of polarity determinants appears to have been serially utilized in both sets of structures. The differences in the precise roles and interactions of the determinants in each structure may represent differences dating from their origin, or alternatively, differences arising from subsequent structural diversifications.

Experimental procedures

Plant material and cultivation

Arabidopsis plants were grown under long-day conditions as previously described (McAbee et al., 2006). Unless otherwise stated, the alleles used in this study were ats-3 (McAbee et al., 2006), phb-6, phv-5 and rev-9 (Emery et al., 2003), and cna-2 (Prigge et al., 2005). Seeds for the triple mutant cna-2 phb-13 phv-11 (Prigge et al., 2005) were a kind gift from Steven Clark (MCDB, University of Michigan). phb-6 phv-5 rev-9/+, ANT:LHG4 and OP:ATS seeds were a kind gift from John Bowman (Biological Sciences, Monash University).

To create ats-3 phb-6 phv-5 and ats-3 phb-6 phv-5 rev-9/+ plants, phb-6 phv-5 rev-9/+ pistils were pollinated with ats-3 pollen on four independent plants. Ten F1 plants from each cross were genotyped for ATS/ats-3, PHB/phb-6, PHV/phv-5 and REV/rev-9 alleles using PCR (see Table S1 for a list of primers used in this study). Three of the F1ats-3/+ phb-6/+ phv-5/+ rev-9/+ plants were allowed to self-pollinate to create segregating F2 populations. Approximately 360 Basta-resistant F2 plants (rev-9 carries a Basta resistance marker) were genotyped for ATS/ats-3, PHB/phb-6, PHV/phv-5 and REV/rev-9 alleles using PCR. Seeds from these plants were evaluated for ats-3 or wild-type morphology. From this population, we identified three ats-3 phb-6/+ phv-5 rev-9/+ individuals. F3ats-3 phb-6 phv-5 and ats-3 phb-6 phv-5 rev-9/+ plants (and subsequent generations) were identified by PCR-based genotyping.

To create ats cna phb phv and ats cna phb phv rev/+ plants, cna-2 phb-13 phv-11 pistils were pollinated with ats-3 phb-6 phv-5 rev-9/+ pollen on two plants. Five F1 plants were genotyped for REV/rev-9; three of the five were rev-9/+. These three individuals were further genotyped for ATS/ats-3, CNA/cna-2, phb-13/phb-6 and phv-11/phv-5; all three were confirmed to be ats-3/+ cna-2/+ phb-13/phb-6 phv-11/phv-5. The resulting segregating F2 population (193 individuals) from a self-pollinated ats-3/+ cna-2/+ phb-13/phb-6 phv-11/phv-5 rev-9/+ plant contained 14 sterile plants and 18 plants with an abnormal seedling phenotype (Figure S4): one or two radialized cotyledons, hyper-accumulation of anthocyanins in cotyledons, no shoot apical meristem and swollen hypocotyls. These phenotypes are similar to the previously described phv rev and phb phv rev phenotypes (Prigge et al., 2005), and occurred at a frequency of approximately 1 in 16. We completed genotyping on 96 of the remaining 193 individuals for ATS/ats-3, CNA/cna-2, phb-6, phv-11/phv-5 and REV/rev-9. Among these 96 F2 plants, we identified one ats cna phb-6 phv individual and four ats cna phb-6 phv rev/+ individuals; these individuals were used for phenotypic analyses.

To create an ectopic ATS expression line, we used the pOpL two-component system (Liu and Meinke, 1998; Moore et al., 1998). The ANT:LHG4 line has been described previously (Schoof et al., 2000). The OP:ATS line was created by cloning the ATS cDNA into the 10-OP BJ36 vector (Moore et al., 1998) using the BamHI site as the 5′ site and the HindIII site as the 3′ site; the OP:ATS cassette was then cloned into pMLBART as a NotI fragment and then transformed by the floral-dip method (Clough and Bent, 1998) into Arabidopsis accession Ler (Y. Eshed, Plant Sciences, Weizmann Institute and J. Bowman, Biological Sciences, Monash University, personal communication). We crossed ANT:LHG4 into OP:ATS to generate ANT>>ATS plants. Ovules in the resulting F1 progeny from three different individuals were phenotyped.

DNA extraction and genotyping

Genomic DNA was extracted from Arabidopsis leaf tissue using 2 × CTAB buffer (2% cetyl-trimethyl-ammonium bromide (CTAB), 1.4 m NaCl, 100 mm Tris/HCl pH 8.0, 20 mm EDTA), followed by chloroform extraction and DNA precipitation with isopropanol at room temperature. DNA pellets were washed with 70% ethanol, re-suspended in 100 μl of sterile water and stored at −20°C. Genotyping was performed by PCR with 2.0 μl of genomic DNA in a 25 μl reaction using either GoTaq Master Mix (Promega, or ExTaq DNA polymerase (Takara,


Primers used in this study for genotyping are listed in Table S1.


Cleared whole-mounts clearings were prepared by dissecting ovules from carpels using needles, and clearing for 2–3 days in a couple drops of Hoyer’s solution (7.5% w/v gum arabic, 6 m chloral hydrate, 5% v/v glycerol) under a coverslip as described by Liu and Meinke (1998). Ovules were photographed on a Zeiss Axioplan microscope ( with Normarski optics using a Zeiss Axiophot camera.

For scanning electron micrography, tissue was fixed and critical point-dried as previously described (McAbee et al., 2006), and imaged on a Philips XL 30 scanning electron microscope (FEI Company,

Light micrographs were taken on a Kodak DC290 camera ( mounted on a Zeiss SV8 stereomicroscope.

All images were acquired digitally and edited in Adobe Photoshop CS2 (

In situ hybridization

Digoxigenin (DIG)-labeled antisense probes for in situ hybridization were synthesized using plasmids purified using a Qiagen miniprep kit ( as previously described (McAbee et al., 2006). For the antisense PHB probe, pGEM-PHB (Williams et al., 2005) was linearized with SphI and transcribed using SP6 RNA polymerase (Promega). For the antisense PHV probe, pGEM-PHV (Williams et al., 2005) was linearized with NotI and transcribed using T7 RNA polymerase (Promega). For the antisense CNA probe, pGEM-CNA (Williams et al., 2005) was linearized with SphI and transcribed using SP6 RNA polymerase. For the antisense INO probe, pJMV86 (Villanueva et al., 1999) was linearized with XhoI and transcribed using T7 RNA polymerase. For the antisense ATS probe, pBS-KAN4 (a gift from John Bowman) was linearized with HindIII and transcribed using T7 RNA polymerase. Tissue fixation and in situ hybridization were performed as described previously (McAbee et al., 2006), with the following modifications: inflorescences were fixed in FAA (3.7% formaldehyde, 5% acetic acid, 50% ethanol) for 2 h at room temperature prior to dehydration in ethanol and embedding in Paraplast-Xtra (Fisher Scientific, Slides were hybridized with approximately 10 pg of DIG-labeled probe overnight at 53°C. Immunological detection of the DIG-labeled probes was performed using a DIG Nucleic Acid Detection Kit (Roche, according to the manufacturer’s instructions. Following detection, slides were rinsed in sterile water and mounted with Crystal Mount (Fisher Scientific) and a coverslip prior to being photographed under Differential Interface Contrast (DIC) using a Zeiss Axiophot camera attached to a Zeiss Axioplan microscope. Digital images were edited using Adobe Photoshop CS2.

Seed set measurements

Measurements of seed number per silique represent the mean values from three individual plants (biological replicates). Each biological replicate comprises the mean value obtained from five siliques (technical replicates). For statistical analysis of genetic effects, the seed number per silique was compared by Kruskal–Wallis one-way anova on ranks, with pairwise comparisons (Student–Newman–Keuls Method), using sigma stat version 3.5 (Systat,


The authors would like to thank John Bowman (Biological Sciences, Monash University), Jennifer Fletcher (PGEC, University of California, Berkeley) and Steven Clark (MCDB, University of Michigan) for generously sharing seeds and plasmids, and Justin Walley, Alexandra Arreola and Quynh Do for technical assistance. We would also like to acknowledge John Bowman, Simon Chan (Plant Biology, University of California, Davis), Jim Doyle (Evolution and Ecology, University of California, Davis) and members of the Gasser laboratory for valuable discussion. This work was supported by National Science Foundation grant IOS0419531 (to CSG) and National Institutes of Health training grant T32M070377 (to DRK).