Modifications of a conserved regulatory network involving INDEHISCENT controls multiple aspects of reproductive tissue development in Arabidopsis


Author for correspondence:

Steve Swain

Tel: +61 2 6246 4814



  • Disrupting pollen tube growth and fertilization in Arabidopsis plants leads to reduced seed set and silique size, providing a powerful genetic system with which to identify genes with important roles in plant fertility.
  • A transgenic Arabidopsis line with reduced pollen tube growth, seed set and silique growth was used as the progenitor in a genetic screen to isolate suppressors with increased seed set and silique size.
  • This screen generated a new allele of INDEHISCENT (IND), a gene originally identified by its role in valve margin development and silique dehiscence (pod shatter). IND forms part of a regulatory network that involves several other transcriptional regulators and involves the plant hormones GA and auxin. Using GA and auxin mutants that alter various aspects of reproductive development, we have identified novel roles for IND, its paralogue HECATE3, and the MADS box proteins SHATTERPROOF1/2 in flower and fruit development.
  • These results suggest that modified forms of the regulatory network originally described for the Arabidopsis valve margin, which include these genes and/or their recently evolved paralogs, function in multiple components of GA/auxin-regulated reproductive development.


Plant reproduction can be divided into the sequential processes of flowering, floral development, pollination, fertilization, seed set and fruit growth. Two major regulators of these processes are the classic plant hormones, auxin and GA, which have both independent and overlapping roles throughout plant development (Hedden, 2003; Vanneste & Friml, 2009; Depuydt & Hardtke, 2011). Auxin regulates a wide range of processes from embryogenesis through to senescence, including maturation of flowers, fertilization and fruit development (Okada et al., 1991; Nemhauser et al., 2000; Benjamins et al., 2001; Ozga et al., 2002; Nagpal et al., 2005; Aloni et al., 2006; Cheng et al., 2006, 2007; Cecchetti et al., 2008; Larsson et al., 2008). Auxin signalling is mediated by the interaction of auxin response factors (ARFs) and auxin/indole acetic acid (aux/IAA) proteins (Guilfoyle & Hagen, 2007). Two partially redundant ARFs, ARF6 and ARF8, function together to promote flower maturation, regulating the development of both maternal and paternal reproductive tissues (Vivian-Smith et al., 2001; Nagpal et al., 2005; Wu et al., 2006b). In addition, ARF8 also restricts premature ovary growth before fertilization (Vivian-Smith et al., 2001; Goetz et al., 2006, 2007).

Gibberellins also regulate many developmental processes, including the growth and function of various reproductive tissues (Swain & Singh, 2005). Endogenous GA concentrations are controlled by a complex but well-characterized biosynthesis pathway (Yamaguchi, 2008), with bioactive GAs promoting the SLEEPY1 (SLY1)-mediated degradation of the growth suppressive DELLA proteins through binding of the GA receptor (Sun, 2011). Arabidopsis has five DELLAs, GA INSENSITIVE (GAI), REPRESSOR OF ga1-3 (RGA), RGA-LIKE1 (RGL1), RGL2 and RGL3, which collectively function in numerous stages of plant development in partially overlapping domains (Itoh et al., 2003; McGinnis et al., 2003; Dill et al., 2004; Gomi et al., 2004; Ueguchi-Tanaka et al., 2005, 2007; Griffiths et al., 2006). Mutants with severe reductions in endogenous GA concentrations produce poorly developed flowers that fail to set seed, predominantly because of defects in stamen development, which include reduced filament elongation, inhibition of anther dehiscence and a lack of mature pollen production (Hu et al., 2008). Mutants with milder GA deficiency, such as 3ox1-1 (defective in one of four Arabidopsis GA3OX genes), or impaired GA signalling, such as gai-1, have flowers that appear more like the wildtype (WT) yet still show reduced fertility, seed set and silique size (Swain & Singh, 2005). These more subtle defects can be attributed to GA promotion of pollen tube elongation, seed development and silique elongation following seed set (Hu et al., 2008).

Because GAs and auxin regulate multiple components of flower development and fertility, their actions must be tightly regulated and coordinated, most likely being governed by complex direct and indirect interactions between transcriptional regulators. An example of this type of hormonal regulation, which involves crosstalk between GA and auxin action, has recently been described for the INDEHISCENT (IND), ALCATRAZ (ALC) and SPATULA (SPT) genes (Sorefan et al., 2009; Arnaud et al., 2010; Girin et al., 2011; Groszmann et al., 2011) in the silique valve margin (Fig. 1). IND, ALC and SPT are basic helix–loop–helix (bHLH) transcription factors that are required for proper valve margin (VM) development and later differentiation of the silique dehiscence zone (DZ), allowing seed dispersal (Rajani & Sundaresan, 2001; Liljegren et al., 2004; Girin et al., 2010, 2011; Groszmann et al., 2011). IND acts upstream of SPT and ALC (Groszmann et al. ; 2011; Girin et al., 2011) and causes localized changes in concentrations and activity of GA through the regulation of GA3OX1 (Arnaud et al., 2010). Changes in GA concentrations lead to DELLA degradation and the release of SPT and ALC from the antagonistic DELLA complex, allowing them to interact and exert transcriptional regulation of their targets (Arnaud et al., 2010; Gallego-Bartolome et al., 2010). The combined action of SPT and IND directs the distribution and therefore localized concentrations of auxin via transcriptional regulation of PINOID (PID) and WAG2 (Sorefan et al., 2009; Girin et al., 2011). Collectively, these interactions lead to an auxin minimum associated with the VM/DZ to ensure correct differentiation and development of this structure. IND, ALC and SPT are also part of a larger regulatory network directing the development of the VM/DZ and adjacent tissues that include FRUITFUL (FUL), REPLUMLESS (RPL), and the SHATTERPROOF (SHP1/2) genes (Ostergaard, 2009; Arnaud et al., 2010; Groszmann et al., 2010, 2011; Girin et al., 2011). However, it is not yet known to what extent, if any, this model (Fig. 1) may apply to other tissues, although associations between SPT and DELLAs are known to have a role in cotyledon and leaf growth (Ichihashi et al., 2010; Josse et al., 2011).

Figure 1.

INDEHISCENT (IND) is part of a regulatory network that specifies valve margin (VM)/dehiscence zone (DZ) identity in Arabidopsis thaliana. The current model of the VM/DZ regulatory network, including several recent observations that provide additional evidence of the members and their interactions (Ostergaard, 2009; Sorefan et al., 2009; Arnaud et al., 2010; Gallego-Bartolome et al., 2010; Groszmann et al., 2010, 2011; Girin et al., 2011). Blue lines represent transcriptional regulation (ALCATRAZ (ALC) regulation by IND is likely but not confirmed); red lines represent protein–protein interactions, effects on protein localization or enzyme activity (GA3OX1); and black lines represent developmental outcomes (VM/DZ development). Auxin action would probably involve the activity of AUXIN RESPONSE FACTOR (ARF) proteins.

In previous work to explore the roles of GAs in reproductive development, a transgenic line ectopically expressing a pea (Pisum sativum L.) cDNA encoding a GA degrading enzyme (PsGA 2-OXIDASE2) was characterized (Singh et al., 2002). This line, 35S:Ps2OX2/28c (hereafter 35S:2ox), has reduced pollen tube growth and consequently poor seed set and short siliques relative to self-pollinated WT plants (Cox & Swain, 2006). These 35S:2ox phenotypes are explained solely by the observed defects in pollen tube growth and can all be rescued by pollination with WT pollen, or by combining the 35S:2ox line with mutants possessing increased GA signalling (Swain et al., 2004). To identify genes with important roles in plant fertility, this 35S:2ox line was used as the progenitor in a genetic screen to isolate suppressors with increased seed set and silique size. This screen identified a new IND allele, and based on this finding we have identified and characterized novel roles for IND, and the closely related HECATE3 (HEC3) gene, in GA and/or auxin-promoted reproduction in Arabidopsis thaliana. In turn, these observations suggest that a regulatory network first described for VM/DZ development, and which includes IND, also functions in additional floral tissues.

Materials and Methods

Growth conditions

Plants of Arabidopsis thaliana (L.) Heynh were grown at 22–24°C under an 18 h photoperiod (white fluorescent light; 60–70 μmol m−2 s−1). Mutagenesis was conducted using standard approaches (see Supporting Information, Methods S1).

Pollen assays and beta-glucuronidase (GUS) staining

Pollen counts were estimated using one healthy medial anther, at the point of dehiscence, from at least 10 individual flowers per genotype (see Methods S1). Pollen viability was examined using fluorescein diacetate, as described by Heslop-Harrison & Heslop-Harrison, 1970 (Methods S1). Pollen tube growth within a pistil was visualized using either aniline blue or X-gluc (when pollinated with LAT52:GUS pollen) stains (Methods S1). Pollen tube lengths were measured from the style to the point in the ovary where the majority of pollen tubes had extended. For GUS staining, samples were harvested into 90% acetone solution, washed in phosphate buffer, transferred into a GUS staining solution, incubated at 37°C for 16 h and then cleared with 70% ethanol ready for visualization (Methods S1).

Scoring seed set and calculating fertility

Twelve to 18 mature siliques per genotype were scored for seed set and silique length and at least two biological replicates were scored for each genotype. Seed set was calculated by removing valves and scoring the number of seeds and/or ovules per silique. Ovules that had been fertilized and subsequently aborted before complete maturation were scored as seeds (this occurred infrequently). Since ind mutants did not differ from WT in the number of ovules per ovary (data not shown), fertility or % seed set was determined by calculating the frequency of successfully fertilized ovules. Unless otherwise stated, data are from self-pollinated flowers.

Statistical analyses

All statistical analyses were carried out using SPSS 14.0 for Windows® statistical package (Microsoft Corporation). Percentage data (e.g. seed set per silique) was transformed to reduce variance heterogeneity. Variation was identified by independent samples T-test or univariate analysis of data and significant differences between means were identified using Tukey's HSD (P ≤ 0.05). Histograms labelled with the same letter are not statistically different from each other, whereas histograms labelled with different letters are (P ≤ 0.05). All error bars represent SEs of the mean and are shown where appropriate.

Hormone analysis

Endogenous concentrations of GA9, GA4 and IAA in entire flower buds were determined using di-deuterated (GAs) or 13C6 (auxin) internal standards and standard methods (Wolbang et al., 2004). Flower buds just before anthesis were pinched off immediately below the receptacle and snap-frozen in liquid nitrogen. Frozen flower buds were stored at –80°C until at least 80 mg of tissue (FW) was harvested.


Loss of IND function improves seed set and fruit size in 35S:2ox plants

35S:2ox M1 plants (c. 3000) were screened for sectors producing one or more larger siliques compared with self-pollinated 35S:2ox plants, with multiple candidate suppressor lines being identified. Assessment of derived M2 populations identified one suppressor line with progeny that, in addition to an increase in seed set and longer siliques (Fig. 2a), also displayed an indehiscent pod phenotype. In three subsequent backcrosses between this suppressor line and the 35S:2ox progenitor, the indehiscent phenotype was consistently found to cosegregate with improved seed set and larger siliques (data not shown). Therefore, this trait was used to map the mutant allele in a cross between the suppressor line (in Ler) and WT Col plants. Using bulk segregant analysis combined with ecotype-specific sequence markers between Col and Ler (Lukowitz et al., 2000), the mutation in the indehiscent suppressor line was mapped to IND, a bHLH transcription factor (Liljegren et al., 2004). The allele isolated in the 35S:2ox suppressor screen was named ind-11 and has a guanine-to-adenine nucleotide change resulting in an arginine (131) to glutamine (131) substitution in the bHLH domain (Fig. 2b). This arginine is conserved in 86% of Arabidopsis bHLH proteins, including the close relatives HEC1, 2 and 3, and confers binding ability to G-box DNA regulatory elements (Toledo-Ortiz et al., 2003), including the CGCGTG variant recognized by IND (Girin et al., 2011).

Figure 2.

indehiscent (ind) acts in both the pollen and pistil to increase 35S:2ox seed set in Arabidopsis thaliana. (a) Typical self-pollinated siliques from wildtype (WT), 35S:2ox, 35S:2ox ind-11 and 35S:2ox ind-12, showing partial rescue of the poor seed set and reduced silique growth in 35S:2ox plants by reduced IND function. (b) Schematic diagram of the IND gene indicating the site of the ind-11 missense mutation and the insertion points of several T-DNA alleles. (c) Seed set as a percentage of total ovules per silique after self-pollination of the indicated genotypes. (d) Seed set from controlled pollinations to investigate the role of IND in the pollen and pistil. bHLH, basic helix–loop–helix; G → A, guanine-to-adenine nucleotide change. Error bars represent standard error of the mean.

The Arabidopsis insertion databases possessed several other ind alleles (Fig. 2b) in addition to those previously identified through their pod shatter defect. To confirm that reduced IND function was responsible for both the indehiscent pod and improved seed set and silique size in the 35S:2ox background, the original 35S:2ox line was crossed to an independent enhancer trap ind allele in the Ler background (ind-12). Consistent with a role for IND in controlling seed set, ind-12 also improved fertility and silique size of 35S:2ox plants (Fig. 2c). Relative to other ind alleles, and consistent with the mutation in a highly conserved residue of the bHLH domain, ind-11 appeared to be a severe and possibly null allele.

IND acts in both the pollen and the pistil

The reduced fertility and silique size of the 35S:2ox line used here are the result of reduced pollen tube growth with no detectable defects in maternal components of fertility or seed development (Singh et al., 2002; Swain et al., 2004). However, despite only a pollen-specific effect of 35S:2ox, reciprocal hand-pollinations indicated that reduced IND function in both the pollen and pistil was required to clearly improve seed set in 35S:2ox plants (Fig. 2d), although subtle effects occur in both tissues independently (see Fig. 3 below). The effect of ind in maternal tissues did not include a change in ovule number (data not shown).

SHP1 and SHP2 are MADS box proteins that act upstream and in parallel with IND to specify the cells of the valve margin layer (Ferrandiz et al., 2000; Liljegren et al., 2000, 2004). Both shp1 and shp2, either alone or in combination, were able to partially rescue the poor fertility of 35S:2ox (Fig. 3a). Reciprocal hand-pollination experiments revealed that the rescue of the poor fertility of 35S:2ox by shp1 and shp2 occurred through the pollen (data not shown; Fig. 3c). These results suggest that both SHP genes function in GA-mediated pollen tube growth, but unlike their redundant action in the VM, SHP1 and SHP2 do not appear to act redundantly in pollen. Consistent with a role in pollen tube growth, both SHP genes are expressed in mature pollen (Schmid et al., 2005).

Figure 3.

SHATTERPROOF1/2 (SHP1 and SHP2) but not ALCATRAZ (ALC) regulate seed set in 35S:2ox plants of Arabidopsis thaliana. (a, b, d) Seed set per silique after self-pollination of indicated genotypes. (c) Seed set from controlled pollinations to investigate the role of INDEHISCENT (IND) and SHP2 in the pollen and pistil. The ind-11 allele was used in (b–d). WT, wildtype. Error bars represent standard error of the mean.

The shp2 allele enhanced the effect of ind on fertility, and the seed set of self-pollinated 35S:2ox ind shp2 mutants was similar to that of WT plants (Fig. 3b). This synergistic interaction allowed the action of IND in pollen tubes and pistil tissues to be examined in more detail. Comparison between WT and ind pistils pollinated with 35S:2ox shp2 pollen confirmed that ind acts in the pistil to improve fertility in the presence of pollen carrying the 35S:2ox transgene. Similarly, comparison between WT pistils pollinated with 35S:2ox shp2 and 35S:2ox ind shp2 pollen confirmed that ind also acts in pollen to improve seed set (Fig. 3c).

ALCATRAZ (ALC) acts downstream of IND in the specification of the DZ (Rajani & Sundaresan, 2001). The alc-1 allele did not detectably improve the poor fertility of 35S:2ox or enhance the partial rescue of 35S:2ox fertility by ind (Fig. 3d).

IND is expressed in developing flowers and siliques

Quantitative reverse transcription polymerase chain reaction (RT-PCR) analysis of IND expression revealed the highest expression in floral tissues just before anthesis and in young siliques 1–2 d after pollination (Fig. 4a). Expression levels decreased in developing siliques by c. 50% by 7 d after pollination. Low levels of IND expression were seen in mature siliques, mature seeds and rosette leaves, with the lowest levels in roots (Fig. 4a). IND expression was also monitored using a GUS insertion line present in the ind-6 gene trap allele (Wu et al., 2006a). GUS expression was detected in developing anthers and pollen, as well as in the silique DZ (Fig. 4b). Expression was visible in the locules of the anthers from around stage 8, becoming restricted to the pollen grains around stage 12 (Fig. 4b, inset). Expression was also detected within the VM around stage 11 and persisted throughout silique development, becoming localized to the lignified layer of the DZ as it differentiated during stage 17 (Wu et al., 2006a). The observed expression pattern of IND is consistent with its reported role in DZ formation (Liljegren et al., 2004; Wu et al., 2006a), the roles in pollen tubes and pistils described earlier, as well as during other stages of flower development.

Figure 4.

INDEHISCENT (IND) is expressed in developing flowers and siliques of Arabidopsis thaliana (a) Quantitative RT-PCR expression of IND in different tissues (relative to the expression in roots). Values are the mean ± SE of three biological replicates. (b) GUS staining from a homozygous plant carrying the ind-6 gene trap line. Inset, anthers at stage 12 with GUS expression in the pollen grains. Error bars represent standard error of the mean.

Reduced IND function does not rescue the pollen tube growth defect of vgd1 and pop2

To investigate whether IND function extends to other pathways required for pollen tube growth and fertilization, the interaction between ind and two mutants with fertility defects thought to be unrelated to GAs was examined. Mutants of VANGUARD1 (VGD1), a pectin methylesterase, have impaired pollen tube growth through the transmitting tract (Jiang et al., 2005), similar to self-pollinated 35S:2ox plants. Mutants of POLLEN-PISTIL INCOMPATIBILITY2 (POP2), a γ-aminobutyric acid (GABA) transaminase, have disrupted pollen tube guidance towards ovules and are almost completely sterile (Palanivelu et al., 2003). The inability of ind-11 to improve the poor fertility of either vgd1 or pop2 (Fig. S1) suggests that IND's role in fertility is distinct from the VGD1 and POP2 pathways.

IND regulates GA-promoted anther and pollen development

Because reduced IND function improved the poor seed set of the 35S:2ox transgenic line, the ability of ind to improve the fertility of a range of mutants with mild or severe defects in GA biosynthesis or signalling was investigated.

Compared with WT plants, anther development is abnormal in the GA-deficient ga3ox1-1 (3ox1) mutant, resulting in reduced pollen production and pollen viability (Fig. 5a–c), associated with a reduction in seed set (Fig. 5d). Loss of RGL2 DELLA protein function increased pollen production, pollen viability and seed set in the 3ox1 rgl2 mutant (Fig. 5a–d). Similarly, loss of IND function also improved pollen production, pollen viability and seed set in the 3ox1 ind mutant (Fig. 5a–d), consistent with IND expression in anthers and pollen (Fig. 4). In contrast to its effects on fertility, ind did not suppress the 3ox1 dwarf phenotype (Fig. 5e).

Figure 5.

indehiscent (ind) suppresses the fertility defects of mild GA mutants of Arabidopsis thaliana (a) Representative anthers at anthesis of WT, 3ox1 and other genotypes: all genotypes can produce pollen, but the amount and viability vary. (b–e) Pollen grains per anther (b), pollen viability (c), seed set (d) and average internode length until the first flower (e) of wildtype (WT), 3ox1 and other genotypes. (f) Representative anthers at anthesis of WT, gai-1 and other genotypes: all genotypes can produce pollen, but the amount and viability vary. (g–i) Pollen grains per anther (g), pollen viability (h) and seed set (i) of WT, gai-1 and other genotypes. (j) Seed set of WT, sly1-10 and other genotypes. Error bars represent standard error of the mean.

The partially improved fertility of the 3ox1 ind mutant and the role of IND in regulating GA3OX1 expression in the VM (Arnaud et al., 2010) raise the possibility that ind alters endogenous GA concentrations in multiple floral tissues. One possibility is that expression of the related GA3OX3 gene, also required for normal anther development (Hu et al., 2008), increases in ind to compensate for the reduced GA3OX1 function. However, no change in GA3OX3 expression was observed in ind anthers (data not shown). In addition, there was no detectable difference in the concentrations of GA4 and GA9 (the major active GA in Arabidopsis and its inactive precursor) in whole flower buds between WT and ind that could explain the partial rescue of 3ox1 (Table S1).

The lack of a clear change in GA concentrations suggests that IND probably has other GA-related roles during flower development that explain the rescue of the 35S:2ox and 3ox1 fertility defects. To test if IND functions downstream of GA metabolism, the ability of ind to improve the fertility of mutants with reduced GA signalling was investigated.

The gai-1 allele encodes a mutant form of a growth-inhibiting DELLA protein that is resistant to GA-promoted degradation (Peng et al., 1997). gai-1 plants have reduced GA response, milder growth defects compared with other GA signalling mutants (discussed later), and under the conditions used here are highly infertile and set few seeds (Swain et al., 2004; Fig. 5f–i). Like 3ox1, gai-1 anthers produced less pollen, and this defect was reduced in gai-1 rgl2 anthers (Fig. 5f–h). Reduced IND function also improved pollen grain numbers and viability in gai-1 ind anthers (Fig. 5f–h), leading to improved seed set of gai-1 ind plants (Fig. 5i). In particular, ind improved the development of gai-1 anthers in early flowers which were otherwise almost completely lacking in pollen (Fig. 5f).

The ability of ind to improve seed set in plants with more severe defects in GA signalling was also examined using two genotypes with strongly elevated amounts of DELLA proteins. The RGA:Δ17RGA-GFP transgenic line expresses and accumulates a stabilized RGA-GFP protein that is resistant to GA-promoted degradation as a result of a deletion in the DELLA domain (Dill et al. 2001). sly1-10 plants are defective in the ability to degrade DELLA protein and accumulate high amounts of RGA and GAI, and presumably other DELLA proteins (Dill et al., 2004). Under standard growth conditions, both RGA:Δ17RGA-GFP and sly1-10 plants have a more extreme phenotype than 3ox1 or gai-1 plants, being severely dwarfed and completely sterile, in part because of defective anther development. In contrast to the effect of rgl2, ind was unable to improve seed set in sly1-10 plants detectably (Fig. 5j) or in plants homozygous for the RGA:Δ17RGA-GFP transgene (no seeds with or without ind present).

These results suggest that reduced IND activity is able to partially suppress the fertility defects of mild but not severe GA mutants. IND appears to affect a GA-related process that is downstream of bioactive GA synthesis, whereby a loss of IND function promotes DELLA degradation or decreases downstream DELLA activity to a degree sufficient to alleviate the fertility defects in mild but not severe GA biosynthetic and signalling mutants. This conclusion is consistent with the similarity between the rgl2 and ind rescue of the 3ox1 and gai-1 fertility defects, and also the ability of rgl2 to improve pollen tube growth and seed set in the 35S:2ox background similar to that of 35S:2ox ind (S.M. Swain, unpublished).

IND regulates auxin-dependent transmitting tract function

Like GA, auxin also has important roles in flower, seed and fruit development. For example, the ARF6 and ARF8 genes, which are part of the auxin signalling pathway, have partially redundant roles in flower development (Nagpal et al., 2005). In particular, the reduced seed set of arf8 plants resembles that of 35S:2ox, but is caused by reduced female fertility rather than an intrinsic defect in pollen tube growth (Vivian-Smith et al., 2001; Swain et al., 2004). To investigate the interaction between IND and ARF8 in regulating female fertility, arf8-4, arf8-4 ind-12 and arf8-4 ind-6, in the Ler background, were pollinated with WT Ler pollen. Both ind alleles partially rescued the poor female fertility of arf8 (Fig. 6a). Similarly, in the Col background, ind-13 markedly improved the fertility of arf8-6 in self-pollinated flowers (Fig. 6b). Thus, in addition to its role in GA action during reproductive development, IND also has a role in auxin-dependent female fertility.

Figure 6.

indehiscent (ind) partially suppresses the arf8 reduced fertility defect in Arabidopsis thaliana. (a) Seed set from controlled pollinations, with wildtype (WT) Ler pollen, to investigate the interaction between ind and arf8 in pistils. (b) Seed set of WT, arf8, arf8 ind and ind. (c) Growth of WT Ler pollen tubes through pistils of different genotypes (monitored using aniline blue staining) after controlled pollinations with WT pollen. st, style. (d) Growth of WT (Col) pollen tubes through pistils of different genotypes (monitored using GUS staining) 18 h after controlled pollinations with Lat52:GUS pollen. (e) Seed set along siliques from different genotypes after controlled pollinations with WT Ler pollen. Ovule position 1 was closest to the stigma. (f) ind enhances the carpel fusion defect of spatula (spt). Error bars represent standard error of the mean.

Because of the genetic interaction between IND and ARF8, and the fact that pollen tubes elongate poorly in the arf6 arf8 double mutant (Wu et al., 2006b), the female fertility defect of the arf8 single mutant alone was examined in more detail using pollinations with WT pollen. The arf8 mutant showed some disruption in transmitting tract function (Fig. 6c,d) in addition to the reported defects in ovule morphology (Vivian-Smith et al., 2001). Compared with Ler, faster-growing Col pollen tubes (Swain et al., 2004) were able to fertilize more ovules in arf8 pistils (Fig. S2), confirming that reduced pollen tube growth, in part, contributes to poor seed set in arf8. Closer examination of the ind rescue of arf8 seed set showed that the poor growth of WT pollen tubes in arf8 pistils was improved by ind in three separate experiments (Fig. 6c–e). These experiments also revealed that reduced IND activity in the pistil can slightly impair WT pollen tube growth (Fig. 6c–d). Based on these results, at least part of IND's role in female fertility appears to be via effects on auxin-dependent transmitting tract development. This function appears to be upstream of ARF8, as IND expression is not altered in the arf8 background (Fig. S3a).

Further evidence of a role for IND in transmitting tract development is provided by the ind spt double mutant (Fig. 6f) in which the severity of spt defects in carpel margin fusion and subsequent transmitting tract development (Alvarez & Smyth, 2002; Groszmann et al., 2008, 2011) are enhanced with the concurrent loss of ind (Fig. 6f; Girin et al., 2011). The role for IND in these processes is likely hidden through partial redundancy with its closest relatives, the HEC genes, which function in transmitting tract development (Gremski et al., 2007). Although SHP1/2 are also involved in carpel fusion and transmitting tract development (Colombo et al., 2010), shp1 and shp2 were unable to improve the poor fertility of arf8 (data not shown) or improve pollen tube growth through the arf8 transmitting tract when WT pollen was applied to arf8-4 shp1 shp2 pistils (Fig. 6d).

As ARF6 and ARF8 have partially redundant functions, the interaction between arf6, the arf6 arf8 double mutant and ind was also examined. Similar to arf6 arf8, flower buds on arf6 arf8 ind triple mutant plants remained underdeveloped (Fig. 7a) and female-sterile even when manually pollinated with WT Col pollen (0% seed set for both genotypes). Loss of IND function also failed to improve the reduced fertility of the arf6 single mutant (Fig. 7b).

Figure 7.

Interactions between auxin response factor6 (arf6), arf8, indehiscent (ind) and hecate3 (hec3) in Arabidopsis thaliana. (a) Representative inflorescences from WT, arf6 arf8 and arf6 arf8 ind plants. Flower development fails to proceed past the stage shown in both mutants. (b) Seed set of wildtype (WT), arf6, arf6 ind and ind. (c) Seed set from controlled pollinations to investigate the role of hec3 in the pollen and pistil. (d) Representative flowers from WT and hec3 arf8 early and late in flower development showing reduced stamen filament elongation (arrowheads) and the absence of pollen in the double mutant. Flowers of hec3 and arf8 single mutants are very similar to the WT, with comparable anther elongation and pollen production. The reduced stamen filament length of the hec3 arf8 mutant is comparable in length to that reported for the arf6 arf8 double mutant (Wu et al., 2006b). Error bars represent standard error of the mean.

Collectively, these experiments reveal that IND has multiple roles in the development and function of specific male and female flower organs during both GA- and auxin-dependent reproductive development.

IND recently diverged from a HEC-like ancestor within the Brassicaceae lineage

Amongst Arabidopsis bHLH proteins, IND is most closely related to the HEC proteins (Bailey et al., 2003; Heim et al., 2003; Toledo-Ortiz et al., 2003). Among these four proteins in Arabidopsis, HEC1 and HEC2 are the most similar to each other, while IND is most closely related to HEC3 (Fig. 8). Using AtIND as a query to find homologous sequences in databases across 20 species produced a phylogeny with two distinct clades (Fig. 8). The HEC1 clade contains orthologues from a variety of angiosperm species, including the Arabidopsis AtHEC1 and AtHEC2 proteins which are products of a recent genome duplication (Blanc et al., 2003). Members of the HEC3 clade are also well represented across the angiosperms and share distinct sequence differences from HEC1 members in the basic region of the atypical bHLH domain, in the HEC and C-terminal domains, and the presence of an acidic domain immediately N-terminal of the HEC domain (Fig. S4).This acidic domain is likely to be involved in transcriptional activation, as it is the only conserved sequence in a region of AtHEC3 and AtIND capable of triggering transcription in a yeast assay (Liljegren et al., 2004; Gremski et al., 2007). Consistent with this conclusion, AtHEC1 and AtHEC2 lack the acidic domain and are unable to activate transcription in yeast assays (Gremski et al., 2007).

Figure 8.

INDEHISCENT (IND) is a paralogue of HECATE3 (HEC3) that is only present in the Brassicaceae. The phylogeny of IND, HEC1/2/3 and related proteins was identified by searching various publicly available databases. The proteins are divided into two main clades, with putative IND orthologues (blue) all belonging to the HEC3 clade. See Fig. S4 caption for sequence alignment and additional details.

It is likely that IND orthologues are a subclade within the HEC3 clade and are distinct from HEC3 sequences through variations in the HEC domain and the basic and HLH regions of the bHLH domain (Figs 8, S4). All identified IND orthologues are within the Brassicaceae, suggesting that IND has recently diverged from a duplication of an HEC3-like ancestor. Since there is no obvious shared synteny amongst genes surrounding IND and HEC3 (data not shown), IND probably arose through a dispersed duplication event and not through one of the recent whole-genome duplication events associated with the Brassicaceae (Franzke et al., 2011).

The IND and HEC proteins were initially considered to have distinct developmental roles (Liljegren et al., 2004; Gremski et al., 2007). However, results presented here and those recently reported by Girin et al., 2011 reveal a hidden role for ind earlier in carpel development overlapping with the known functions of the HEC proteins. Given this partial redundancy between IND and HEC functions and the apparent recent divergence of IND from HEC3, we decided to investigate hec3 mutants for further defects in fertility.

HEC3 promotes stamen development and pollen tube guidance

Despite the redundancy between the HEC proteins, the hec3 single mutant exhibits mild defects in transmitting tract function associated with retarded pollen tube growth (Gremski et al., 2007). Although a defect in overall WT pollen tube growth through the hec3-1 pistil was not observed under the conditions used here (Fig. 9a), the hec3 mutant still exhibited reduced seed set as a consequence of impaired female fertility (Fig. 7c). The frequency of unfertilized ovules increased along the ovary as the distance from the stigma increased (Fig. 9b), despite pollen tubes growing down the hec3 transmitting tract at the same rate as in WT pistils. Unlike pollen tubes growing within WT ovaries, which grew along funiculi and directly into the ovule micropyle, hec3 ovaries exhibited a defect in the second stage of pollen tube guidance, with pollen tubes often unable to navigate into the mycropyle to fertilize the ovule. After pollen tubes exited the hec3 transmitting tract, they appeared to grow randomly within the ovary, often growing over ovule surfaces and the inner surface of the carpel walls (Fig. 9c). Examination of ovule morphology did not reveal any major defects in hec3, although hec3 seeds are abnormally shaped at maturity with a disrupted seed abscission zone (Ogawa et al., 2009). Interestingly, incorrect patterning of ARF6/8 can also result in misshaped ovules with defective seed abscission zones and a reduced ability to attract pollen tubes (Wu et al., 2006b; Todesco et al., 2010). In terms of female fertility, hec3 was epistatic to ind (Figs 7c, 9b), and IND expression in flower buds just before anthesis was not significantly altered in hec3 (Fig. S5). Thus, HEC3 is required for normal attraction of pollen to ovules and subsequent fertilization and seed set.

Figure 9.

HECATE3 (HEC3) is required for targeting of ovules by pollen tubes and for seed set in Arabidopsis thaliana. (a) Growth of wildtype (WT) Col pollen tubes through pistils of different genotypes (monitored using aniline blue staining) after controlled pollinations with WT pollen. HAP, h after pollination. (b) Seed set along siliques from different genotypes after controlled pollinations with WT Col pollen. Ovule position 1 was closest to the stigma. (c) Representative images of pollen tubes showing correct targeting of the micropyle of a WT Col ovule and a pollen tube that is growing over a hec-3 ovule without entering the micropyle. Error bars represent standard error of the mean.

The function of HEC3 was further explored in the 3ox1 and arf8 backgrounds. While no obvious interactions were observed between 3ox1 and hec3 (data not shown), these experiments did reveal an additional role for HEC3 in stamen development. While hec3 alone did not cause any obvious stamen defects, hec3 exacerbated the abnormal stamen phenotypes previously reported for arf8 and arf6 mutants (Wu et al., 2006b). Filaments in young hec3 arf8 flowers were shorter than filaments in early arf8 flowers, and resembled those of arf6 arf8 flowers (Fig. 7d). Furthermore, hec3 arf8 double mutant anthers were underdeveloped and usually failed to dehisce and release pollen, similar to arf6 arf8 anthers.

Thus, in addition to their roles in transmitting tract function and seed shape/abscission, HEC3 and ARF8 functions overlap in pollen tube guidance to ovules and in stamen (filaments and anther) development.

IND and HEC3 have opposite effects on seed-dependent silique growth

In Arabidopsis there is generally a clear linear relationship between total seed number and final silique length across the full spectrum of seedless siliques to those that are completely filled (Cox & Swain, 2006). Both IND and HEC3 are involved in seed-promoted silique growth, as demonstrated by larger siliques in ind mutants and smaller siliques in hec3, independent of the effects of these genes on seed numbers (Fig. 10a,b). Reduced IND function was able to slightly increase seed-promoted silique size in the hec3 ind double mutant, but could not completely restore growth back to WT values (Fig. 10b). The effects of IND on silique growth are unlikely to be an indirect consequence of defects in VM development since the shp2-1 mutant, in which VM/DZ development is apparently normal, also increased seed-promoted silique growth (Fig 10c). Roles for IND, SHP2 and HEC3 in seed-promoted silique growth are consistent with known roles for both GA (Hu et al., 2008) and auxin (Dorcey et al., 2009) in this process.

Figure 10.

indehiscent (ind) and hecate3 (hec3) alter seed-promoted fruit growth in Arabidopsis thaliana. Silique length as a function of seeds per silique. Lines of best fit are shown for each genotype. Differences in final silique lengths for the genotypes shown are statistically significant (< 0.01). (a) Wildtype (WT) Ler and several ind alleles after controlled pollinations with WT Ler or 35S:2ox pollen to give a range of seed numbers. (b) Self-pollination of WT Col, ind-13, hec3-1 and ind-13 hec3-1. Siliques with a range of seed numbers were selected for analysis. (c) Self-pollination of WT Ler and shatterproof2-1 (shp2-1). Siliques with a range of seed numbers were selected for analysis.


To identify genes with important roles in fertilization and seed set in Arabidopsis, we have used a forward genetic screen on a 35S:2ox transgenic line with reduced pollen tube growth (Singh et al., 2002). In contrast to WT, where almost all ovules are fertilized, the pollen tube growth defect results in the probability that an individual ovule will be fertilized, decreasing from c. 50% at the apical end of the pistil (nearest the stigma) to c. 0% at the base. The reduced seed set means that 35S:2ox siliques are c. 50% shorter than those of WT Ler plants (Cox & Swain, 2006). We anticipated that using 35S:2ox should be a very sensitive approach to identifying genes regulating plant fertility. This mutant screen identified a new allele of IND, based on the subtle but combined effects of reduced IND function on pollen tubes, pistil tissues and seed-promoted silique elongation, all of which contribute to increased silique growth in the 35S:2ox ind background.

INDEHISCENT is a bHLH transcription factor that functions as part of a larger gene network regulating auxin and GA action during VM/DZ development. In this study we define novel roles for IND in multiple additional aspects of GA- and auxin-regulated flower and silique development. Our observations suggest that modified forms of the regulatory network originally described for the Arabidopsis valve margin (Fig 1), which include these genes and/or their recently evolved paralogues, function in multiple components of GA/auxin-regulated reproductive development that predate the evolution of pod shatter in the Brassicaceae.

New roles for IND, HEC3 and SHP1/2 in flower and fruit development

Analysis of proteins that share the highest similarity to IND in Arabidopsis and other species suggests that HEC3 is the closest Arabidopsis paralogue. While HEC3 relatives are present throughout a diverse range of angiosperms, IND sequences could only be identified within the Brassicaceae, suggesting that IND is a relatively young gene that arose from a HEC3-like ancestor near the emergence of the Brassicaceae, some 24–40 million yr ago (Franzke et al., 2011). While IND and HEC expressions overlap in the developing septum and both have a modest role in carpel fusion and transmitting tract development (this study; Girin et al., 2011), the HEC genes are not expressed in the valve margins and have no reported DZ defect (Gremski et al., 2007). Although most gene copies are lost following duplication (Wagner, 1998; Lynch & Conery, 2003), the retention of IND is probably a result of the establishment of a novel VM role (neofunctionalization) distinct from any HEC function. The partial redundancy that remains between IND and HEC genes probably persists to stabilize the development of carpel margins and transmitting tract tissues.

Characterization of ind and hec3 alleles in a range of mutants defective in GA and/or auxin biosynthesis and signalling revealed that IND and HEC3 function in both male and female flower organs. In stamens, IND is involved in another development and dehiscence, pollen formation and subsequent pollen tube growth after pollen germination on the stigma. In female tissues, IND regulates the fusion of the carpel margins and transmitting tract development. HEC3 functions during stamen filament elongation and anther formation, and plays a role in the ability of ovules to attract pollen tubes. Finally, both IND and HEC3 are required for normal seed-dependent silique growth. These results extend the previously described roles for IND in VM/DZ formation (Liljegren et al., 2004), and roles for HEC3 in fusion of the style apex and transmitting tract development (Gremski et al., 2007), as well as in seed shape and the formation of the seed abscission zone (Ogawa et al., 2009).

We have also identified new roles for SHP1 and SHP2 in GA-regulated pollen tube growth and a role for SHP2 in seed-regulated silique growth. In contrast to their roles in VM/DZ development, SHP1 and SHP2 are not fully redundant in pollen or silique growth. Although we could not detect an obvious role for ALC in fertility using our 35S:2ox system, it has recently been reported that ALC functions redundantly with SPT in the development of septum/transmitting tract, style and stigma during gynoecium development (Groszmann et al., 2011).

IND is part of a regulatory network present in multiple tissues

Detailed analysis of VM/DZ development has revealed a regulatory network required for the correct specification and function of this tissue. This network initially included IND, ALC and SHP1/2 and localized changes in GA and auxin concentrations (Ostergaard, 2009; Sorefan et al., 2009; Arnaud et al., 2010; Fig. 1). Recent results have extended this model to include regulation of SPT (and probably ALC) by IND (Groszmann 2010; 2011; Girin et al., 2011), and direct transcriptional regulation of PID and WAG2, two genes involved in auxin movement, by the combined action of SPT and IND (Girin et al., 2011). In addition, SPT, like its paralogue ALC, has recently been shown to interact directly with DELLA proteins and to function during VM/DZ development (Gallego-Bartolome et al., 2010; Groszmann et al. 2011). As a consequence of these interactions, GA concentrations increase and auxin concentrations are reduced in developing VM/DZ cells, which, together with other regulators such as SHP1/2 and auxin signalling components, allows VM/DZ specification (Fig. 1).

A key observation described here is that IND, its closest relative HEC3, and SHP1/2 also regulate GA- and/or auxin-dependent development events in other floral tissues, including anthers, pollen tubes and the septum/stigma/transmitting tract. This regulatory network appears to fine-tune GA action and allow correct organ development and growth; for example, both too little and too much GA have been linked to defective pollen tube growth (see Singh et al., 2002 and references therein). This genetic evidence, together with known gene expression data, implies that the VM/DZ regulatory network, or a modified form involving different paralogues, regulates development of multiple floral tissues (Fig. 11a–b). Furthermore, the observation that HEC3, SEEDSTICK (STK; the closest relative of SHP1/2) and mi167 (targets ARF6/8) are required for correct formation of the seed abscission zone (Pinyopich et al., 2003; Ogawa et al., 2009; Todesco et al., 2010) suggests that a modified form of this network also acts in this tissue (Fig. 11a). Of these different tissues/organs, the silique VM/DZ is the most recently evolved structure, and is unique to the Brassicaceae. Consistent with adaption of this regulatory network to the VM/DZ during evolution of this family, IND and ALC are recently diverged from HEC3 and SPT, respectively, and all four genes retain overlapping roles in development of the false septum/stigma/transmitting tract (this study; Groszmann et al., 2011; Girin et al., 2011). The ability of IND, HEC1/2/3, SPT and ALC to form heterodimers with each other (Gremski et al., 2007; Girin et al., 2011; Groszmann et al., 2011), potentially leading to slightly different regulatory targets, may partly explain how this regulatory network appears to have been successfully adapted to several different tissues.

Figure 11.

INDEHISCENT (IND) is part of a regulatory network present in multiple tissues of Arabidopsis thaliana. (a) Components of the valve margin (VM)/dehiscence zone (DZ) network, and/or their paralogues, are expressed in other floral tissues and may function in a similar manner to regulate development. Auxin action involves AUXIN RESPONSE FACTOR (ARF) proteins (e.g. ARF6/8) and GA action includes GA metabolism (e.g. GA3OX1) and GA signal transduction (e.g. DELLAs). Genetic evidence supports a role for the genes shown in black (this paper and see text) while those in red have only been shown to be expressed in the tissues indicated. TT, transmitting tract. (b) A modified form of the regulatory network active in the VM/DZ functions in anther/pollen/pollen tube development. IND positively regulates SPATULA (SPT) expression, and DELLA, SPT (and possibly ALCATRAZ (ALC) and SHATTERPROOF1/2 (SHP1/2) act as negative regulators of GA action in these tissues. It is also consistent with our results that DELLA partially antagonizes SPT and ALC function, as in the VM/DZ model, to fine-tune development. The simplest explanation for the opposite effects of ind and hec3 on auxin-mediated growth is that patterns of auxin redistribution differ between the two mutants, resulting in differences in localized auxin concentrations. (c) Reduced 3OX1 activity, 35S:2ox or gai-1 can all lead to impaired anther and/or pollen tube development and reduce seed set, which can be rescued to varying degrees by reducing the activity of IND, SHP1 or SHP2. The ind rescue of these mild GA-deficient defects may occur through down-regulation of SPT and similar proteins, which act in addition to the partial antagonism by DELLA of these proteins, to reduce the overall collective suppression on GA-dependent development. The ind rescue may also function through a modest impact on DELLA stability/protein activity (not shown). However, ind is unable to rescue severe GA mutants (i.e. Δ17RGA or sly1) as the suppression of high DELLA activity levels are epistatic to the additional affects on SPT and similar proteins associated with a loss of IND function. Note that the downstream effects of SPT (ALC and related proteins) regulation on GA-dependent anther/pollen/pollen tube development may include, in part, alterations to auxin distribution and localized concentrations as indicated in (b).

Mechanisms of IND/HEC3 action

With regard to the GA pathway, the VM/DZ model (Fig. 1) suggests that IND is a positive regulator of GA action, which acts by promoting GA production, leading to reduced DELLA activity. However, our observations that loss of IND function can improve the fertility of 35S:2ox, 3ox1 and gai-1 imply that IND can also act on and/or downstream of the DELLA proteins as a negative regulator of GA response in some nonVM/DZ tissues, including anthers and pollen. This suggests that the interaction between IND and GA varies in different reproductive tissues. One explanation for changes in IND's influence (positive or negative) on GA action is that downstream components of the IND pathway have modified roles in different tissues. One possible candidate for this role is SPT, which functions as a positive regulator of septum, style, stigma and VM/DZ development (Groszmann et al., 2008, 2011), and as a negative regulator of GA-promoted seed germination and cotyledon, leaf and petal growth in a manner similar to the growth inhibition by the DELLA proteins (Penfield et al., 2005; Ichihashi et al., 2010; Josse et al., 2011). If SPT is also a negative regulator of GA action in anthers/pollen, this could explain the rescue of GA-related defects by ind and the different affects of IND on GA action between VM/DZ tissues and pollen/anthers (Fig 11b). Testing of this hypothesis would ideally require reducing SPT (and possibly ALC) function in anthers and pollen tubes (but not the gynoecium) to determine if seed set is restored in different GA-related mutants.

In terms of IND's role in auxin-regulated reproductive development, it is likely that these relate to localized changes in auxin concentrations in specific reproductive tissues during flower development that are too subtle to detect in whole flower buds (Table S2). This model is consistent with the ability of IND to regulate localized auxin concentration in the VM/DZ (Fig. 1; Sorefan et al., 2009; Girin et al., 2011), and with the observed interactions between ind, hec3, arf6 and arf8. HEC3 is also likely to modify localized auxin concentrations since 35S:IND and 35S:HEC3 cause similar growth defects that can be explained by misdirected auxin movement (Gremski et al., 2007; Sorefan et al., 2009). Thus, IND and HEC3 may regulate auxin concentrations in specific reproductive tissues throughout flower development, including the VM/DZ, as part of a conserved regulatory network. It is possible that changes in localized auxin concentrations also assist in the rescue of the GA fertility defects examined here. One possibility is that auxin destabilizes DELLA proteins, as suggested by Fu & Harberd (2003), although this idea has been disputed (Desgagne-Penix & Sponsel, 2008; Ross et al., 2011; Chapman et al., 2012).

In conclusion, we have identified new roles for IND, HEC3 and SHP1/2 in GA and/or auxin-dependent flower and fruit development, and suggest that they interact with the DELLA pathway to precisely regulate hormone-dependent development. These results extend the known functions of these genes and suggest that modified forms of the regulatory network originally described for the VM/DZ, which include these genes and/or their recently evolved paralogues, function in multiple components of reproductive development.


We thank Angelica Jermakow and Carol Sigston for help with mapping of IND, the CSIRO Summer Student Programme including Stephanie Young, Ellen Sandell, Broderick Matthews and Kylie O'Keefe, the Weeds CRC for financial support to M.G. and La Trobe University and CSIRO for P.K.'s PhD scholarship.