Functional domains of SPATULA, a bHLH transcription factor involved in carpel and fruit development in Arabidopsis

Authors


* (fax 61 3 9905 5613; e-mail david.smyth@sci.monash.edu.au).

Summary

The SPATULA (SPT) gene is involved in generating the septum, style and stigma: specialized tissues that arise from carpel margins. By matching sequences within the extended bHLH region of AtSPT across species databases, twelve orthologues were identified in eudicots, rice and a gymnosperm. Two conserved structural domains were revealed in addition to the bHLH region: an amphipathic helix and an acidic domain. These are conserved in the tomato orthologue, which can restore carpel function to spt mutants of Arabidopsis. The acidic domain is essential for SPT carpel function, and the amphipathic helix supports it. A bipartite sequence overlapping the bHLH domain is required for nuclear localization, and a mutation in the conserved beta strand adjacent to the bHLH C terminus results in the loss of SPT function. SPT apparently acts as a transcriptional activator, as the addition of the SRDX repression domain phenocopies the spt mutant phenotype. Expression of an artificially activating 35S:SPT-VP16 construct can induce carpelloid properties in sepals, and new defects in the gynoecium. These disruptions are associated with ectopic expression of the STYLISH2 gene, although STYLISH2 expression does not require SPT function. Ectopic expression of unmodified SPT does not induce such changes, implying that SPT acts in association with essential coactivators present only in regions where SPT is normally active. Because the VP16 activation domain can compensate to some extent for the loss of the amphipathic helix and acidic domain, these domains may normally interact with such co-activators.

Introduction

Morphogenesis in plants requires the action of a developmental cascade of specific regulatory genes. For example, in Arabidopsis the identity of the most complex organ of flowers, the gynoecium, is defined by the transcription factor gene AGAMOUS (AG), in combination with the SEPALLATA (SEP) genes (Pelaz et al., 2000). Once the gynoecium identity is established, other genes define its differential growth (polarity) and its differentiation. The outer–inner axis of polarity is controlled to some extent by the CRABS CLAW (CRC) gene and its YABBY family relatives (Eshed et al., 1999). On the other hand, apical–basal polarity (stigma, style and ovary) correlates with the apical production of auxin, with the ETTIN (ETT), SPATULA (SPT) and STYLISH (STY) genes among others playing a role in ensuring appropriate differential auxin responses down the developing gynoecium (Nemhauser et al., 2000; Sohlberg et al., 2006).

Lateral–medial polarity is associated with a valve wall developing laterally, and a replum, septum and placentae arising medially. This polarity is predicted by the early action of boundary ge\nes, and genes that maintain the undifferentiated state medially between the two carpels (Alvarez and Smyth, 2002). Subsequently, specific genes are associated with the development of specific tissues from the carpel margins, including the placenta [LEUNIG (LUG) and AINTEGUMENTA (ANT) (Liu et al., 2000)], the transmitting tract within the style and septum [SPT (Alvarez and Smyth, 1999, 2002) and NO TRANSMITTING TRACT (NTT) (Crawford et al., 2007)], and the style itself [STY1 and STY2 (Kuusk et al., 2002)]. In the mature gynoecium and the fruit (silique) that develops from it, genes are active in distinguishing the lateral valve [FRUITFULL (FUL) (Gu et al., 1998)] from the medial replum [REPLUMLESS (RPL) (Roeder et al., 2003)], and in defining the dehiscence zone that allows the separation of the valves and the release of seeds [SHATTERPROOF1 and 2 (SHP1 and SHP2) (Liljegren et al., 2000), ALCATRAZ (ALC) (Rajani and Sundaresan, 2001) and INDEHISCENT (IND) (Liljegren et al., 2004)].

In Arabidopsis the SPT gene plays a role in the development of all tissues that arise from the carpel margins, being required for transmitting tract development and supporting the development of the septum, style and stigma (Alvarez and Smyth, 1999, 2002). It encodes a transcription factor (Heisler et al., 2001) that is a member of the large basic helix-loop-helix (bHLH) family (Bailey et al., 2003; Heim et al., 2003; Toledo-Ortiz et al., 2003). The conserved basic region of this family has been shown to confer specificity in the DNA target that is bound, as well as in some cases to confer a joint function in nuclear localization. The HLH region is associated with homo- or heterodimerization.

Within a species, transcription factor genes often occur in large families (Riechmann et al., 2000). The accumulation of sequence data from a wide range of species has helped in the identification of orthologues within families. Through comparison of such orthologues, conserved structural domains not previously recognized may be revealed. We have adopted this approach and have identified twelve orthologues of the Arabidopsis bHLH gene SPT. Comparison of their sequence has revealed two new conserved structural domains: an amphipathic helix and an acidic domain. By selective truncation and deletion of these domains in transgenes inserted into spt mutant plants, we have assessed their ability to support SPT gynoecium and fruit development function. The amphipathic helix adds to this function, but the acidic domain is required for it. We have also tested whether SPT activates or represses its targets by adding an artificial repressor or activator domain to the SPT protein. SPT is a transcriptional activator, although it apparently requires co-activators for its carpel development function.

Results

Regions conserved in orthologues of SPATULA

A total of 162 bHLH genes have been identified in Arabidopsis (Bailey et al., 2003). Of these, SPT falls into a group of 15 genes with a conserved C-terminal extension of the bHLH region (Heim et al., 2003; Toledo-Ortiz et al., 2003), here called the beta strand. A search for orthologues of the SPT protein was carried out by screening plant genomic and cDNA databases to identify other conserved, and thus potentially functional, domains (Figure S1). A sequence of 62 amino acids was used, including the 49-amino-acid bHLH region, together with the four adjacent N-terminal amino acids that are part of a predicted bipartite nuclear localization sequence (NLS), and the nine residues immediately beside the C terminal of the bHLH region in the beta strand (Figure 1). Homologous genes were considered likely to be SPT orthologues if they matched this 62-amino-acid sequence with closer identity than the closest relative of SPT in Arabidopsis, ALC, which shares 51 out of 62 identical residues (Rajani and Sundaresan, 2001) (Figure S1). SPT and ALC are apparently the products of a recent gene duplication in an ancestor of the Brassicaceae, as sequences more closely related to ALC than to SPT could not be identified outside this family.

Figure 1.

 Alignment of conserved regions with defined secondary structure in the SPATULA protein and its orthologs.
Orthologues of the SPATULA protein were identified in databases based on closest matches to the 62 conserved amino acids of the extended bHLH region (Figure S1). This region includes a bipartite NLS, the overlapping bHLH region proper and the C-terminal beta strand. Two other conserved regions with predicted secondary structure were identified: a putative amphipathic helix and an acidic alpha helical domain. The number of amino acids between conserved regions is shown. The species and source of the sequences were as follows: AtSPT, Arabidopsis thaliana At4g36930; GhSPT, Gossypium hirsutum contig of GenBank DT572940 and DN780013; CcSPT, Citrus clementina GenBank DY289192; FaSPT, Fragaria × ananassa GenBank AY679615; MtSPT, Medicago truncatulaTC97958/GenBank AC144431; GmSPT, Glycine maxTC207452/GenBank AC170860; PtSPT, Populus trichocarpa JGI eugene3.00400325; VvSPT, Vitis vinifera GenBank AM434294; SlSPT, Solanum lycopersicumTC174571; EcSPT, Eschscholzia californica Annette Becker, personal communication; OsbHLH107, Oryza sativa Os02g56140 (TIGR annotation 11972.m10638); OsbHLH108, Oryza sativa Os06g06900 (TIGR annotation 11976.m05415); PgSPT, Picea glaucaTC30545.

The closest matches were found in translated unigenes of other eudicots, including poplar and tomato (60/62 identical), cotton, orange, strawberry, grape and California poppy (59/62 identical), and Medicago truncatula and soybean (58/62) (Figure 1). Among monocots, the two closest relatives of SPT in rice, OsbHLH107 and OsbHLH108 (Li et al., 2006), showed 58 and 59 amino acids identical out of 62, respectively. More distantly, a sequence showing 56/62 identity was found in a database of the white spruce Picea glauca, which is a gymnosperm. Orthologues were not found in the genome of the moss Physcomitrella patens. The two moss bHLH sequences closest to SPT (proteins 164347 and 164436) were identical to SPT for 50 and 51 of the 62 amino acids, respectively, but they were even closer to another member of the SPT group, PIF3 (Figure S1) (identical for 53 and 54 positions).

A second sequence closely related to SPT was identified in tomato, but with slightly lower identity (57/62) (Figure S1). Sequences very similar to this were found in three other species of the Solanaceae, but were not found more widely in the core eudicots sampled. This SPT-S group seems to represent the divergent product of duplication of SPT in an ancestor of the Solanaceae, and is not considered further here.

Alignments of the full sequences of putative SPT orthologues (Figure S1) revealed two further conserved structural domains in eudicot sequences (Figure 1). The first occurs close to the N terminus, and this is likely to fold into an amphipathic helix (Figure 2a) (Heisler et al., 2001). This was not identified in SPT orthologues of rice or spruce (Figure 1), although a related domain is present at a similar N-terminal position in the SPT-S group and in ALC (Figure S1). A second conserved sequence with well-defined secondary structure lies closely upstream of the bHLH domain, and it is characterized by a strongly acidic composition. This is predicted to adopt an alpha helical structure, with the charged acidic residues clustered on one face surrounding a conserved cysteine (Figure 2b). It is present in the strict SPT orthologues of the eudicots, including the basal species Eschscholzia californica, but neither in the monocot or the gymnosperm sampled (Figure 1), nor in ALC. Finally, the nine amino acids downstream of the bHLH are predicted to form a beta strand (Figure 2c), but less conserved sequences further downstream (Figure S1) do not have a clear secondary structure.

Figure 2.

 Helical wheels of the predicted structures of three conserved domains within orthologues of SPATULA.
The amphipathic helix (a) and the acidic domain (b) are drawn as alpha helices, with each successive residue offset by 100°. The beta strand (c) is drawn with angles between successive residues of 160°. Residues are the consensus of the SPT orthologues shown in Figure 1, and occur at each position in at least seven out of the 10 sequences (or 11 out of 13 for the beta strand). D and E, and S and T, are considered equivalent in each case. Other positions are indicated with an asterisk. Non-polar amino acids are in squares; acidic amino acids are circled.

Tomato SPT can support SPT gynoecium function in Arabidopsis

To determine if the regions conserved between the Arabidopsis and tomato (Solanum lycopersicum) SPT proteins were sufficient to support the gynoecium function of SPT, the ability of the tomato protein SlSPT to complement the spt-2 mutant phenotype was tested. The 35S promoter was used as this is capable of restoring the spt-2 mutant silique to wild type when driving expression of an Arabidopsis SPT cDNA (Figure 3a). Of 21 independent 35S:SlSPT transformants, 15 showed some complementation of gynoecium defects when assessed in mature siliques, and the top seven lines were scored for the length of silique, the number of seeds and the percentage of siliques with a fully restored septum (Table 1). The mean value of each variable was much closer to wild type than to the spt-2 mutant, and individual siliques were seen that were very close to the wild type (Figure 3b). Thus, the tomato SPT orthologue can fully supply SPT gynoecium function in Arabidopsis.

Figure 3.

 Complementation of the spt mutant silique phenotype, and nuclear localization of the SPT protein.
(a) Comparison of a mature spt-2 mutant silique (left) and a silique from a spt-2 mutant plant carrying the 35S:SPT transgene (right), showing full restoration of silique length, seed set and septum development (valve removed in each case).
(b) Comparison of an spt-2 mutant silique (left) and a silique from an spt-2 mutant plant carrying the 35S:SlSPT transgene from tomato (right), showing apical fusion and near wild-type length.
(c) Nuclear fluorescence in onion epidermal cells transfected with 35S:SPT-GFP.
(d) Loss of nuclear localization using the same construct carrying a deletion of the SPT nuclear localization sequence [35S:SPT-GFP (ΔNLS)].

Table 1.   Complementation of the silique phenotype of spt-2 mutant plants by the tomato orthologue of SPT
 Wild type (Ler)spt-235S:SlSPT in spt-2
  1. Results are for the first 10 flowers on the main stem in each case, and are shown as mean values ± SEM. The 35S:SlSPT transgenic plants scored were the seven most strongly complemented plants from among 21 independent T1 transformants.

No. plants    557
Mean silique length (mm)  12.03 ± 0.433.60 ± 0.139.01 ± 0.19
Mean no. seeds per silique  56.8 ± 5.91.8 ± 0.431.7 ± 1.6
Mean percentage with full septum100%0%60.0 ± 5.8%

Phenotypes of spt mutants correlate with deduced disruptions to the SPT protein

Next, defects in fruit development of three spt mutants with known mutational changes (Heisler et al., 2001) were compared (to date, only flower developmental defects have been described; Alvarez and Smyth, 1999, 2002). Siliques that developed from the first formed flowers were more severely affected, with later developing fruits becoming progressively closer to the wild type (Figure 4). Two of the mutants (spt-2 and spt-3) were closely similar in their level of disruption, whereas the third (spt-1) was less severe in every aspect. The spt-1 mutant is predicted to result in a stop codon within the amphipathic helix (codon 51) (Figure 5a), so its intermediate phenotype was unexpected, and may result if translation reinitiates at the next available methionine (codon 107) (Heisler et al., 2001). The other two mutants may reflect the null phenotype. The spt-2 mutant results from a conservative arginine to lysine change at codon 209 within the bHLH domain (Figure 5a), and its strong phenotype may result from a loss of DNA binding, as arginine is absolutely conserved at this position in all SPT orthologues (Figure 1), and in more distantly related bHLH proteins (Heisler et al., 2001). The spt-3 mutant is predicted to result in a stop codon at codon 149 before the acidic and bHLH domains (Figure 5a), and the latter are, at least, likely to be essential for function.

Figure 4.

 Properties of the mature fruit of wild type and spt mutant plants.
The mean length of the silique (a) and the mean number of seeds per silique (b) were recorded individually for the first 25 siliques on the main inflorescence. Lines scored are the Landsberg erecta wild type, spt-1, spt-2 and spt-3 mutants, and transgenic spt-2 plants carrying the 35S:SPT construct revealing close to full complementation. Points represent mean values and bars extend for the SEM in either direction, and the number of plants scored in each category is shown.

Figure 5.

 Location of conserved structural domains, and the positions of mutations, truncations and deletions of the SPT protein.
(a) The 373 predicted codons of the SPT protein, showing the positions of the three methionine codons in the N-terminal region (M1–M3), and the conserved structural domains. The codons affected in four point mutations of SPT are also shown.
(b) Regions present in six 35S:SPT truncations and deletions (solid lines) that were transformed into spt-2 mutant plants. The level of complementation observed is shown.
(c) Region present in an N-terminal deletion of SPT, and the consequence of adding the VP16 activation domain to the C terminus.

A new mutant, spt-8, was shown to result from a G-to-A transition in the splice acceptor site (AG) of the fourth intron. This leads to the intron not being spliced from the mRNA (shown by RT-PCR), and is predicted to result in the insertion of 10 different amino acids after codon 251 in the conserved beta strand (Figure 5a), thereby disrupting its predicted secondary structure (Figure 2c). The gynoecium and fruit phenotype of spt-8 is closely similar to the other two strong mutants spt-2 and spt-3 [mean silique length (±SEM) for the first 15 siliques, 3.97 ± 0.12 mm (n = 5); mean number of seeds per silique 0.13 ± 0.10; compare with Figure 4]. This indicates that integrity of the beta strand and/or regions between it and the C terminus are necessary for function. Each of the four mutants was found to be fully recessive in terms of gynoecium development.

Deletion analysis of the SPATULA protein reveals regions that are essential, supportive or dispensable for function

Next, the consequences of deleting subregions of the SPT protein on its ability to complement the loss of SPT function were investigated (Figure 5b). As a starting point, a 35S:SPT construct was used. Quantitatively, this is able to fully complement the gynoecium and silique defects in spt-2 and spt-3 mutant plants (Figure 4).

The first deletion construct [35S:SPT (Met2)] removed the equivalent of the 5′ region of the cDNA so that translation commenced at the predicted second methionine of the open reading frame (codon 16). Compared with the full-length 35S:SPT construct, there was no significant reduction in its ability to complement silique development in spt-2 plants (Figure 6). It should be noted that translation may normally commence at this second predicted methionine. The first methionine is not conserved in SPT orthologues (Figure S1), or in the orthologue of the closely related Brassica oleracea (Figure S2). Also, the nucleotides adjacent to the second predicted methionine (Heisler et al., 2001) are closer to the plant consensus (Joshi et al., 1997; Kozak, 1997) (Figure S2).

Figure 6.

 Degree of restoration of wild-type silique properties in spt-2 mutant plants carrying transgenes encoding truncations and deletions of the SPT protein.
The mean length of the silique (a), the mean number of seeds per silique (b) and the percentage of septum present (c) were recorded individually for the first 15 siliques on the main inflorescence. Lines scored were Landsberg erecta wild type (five plants), spt-2 (five plants) and spt-2 carrying various 35S:SPT constructs [35S:SPT (five plants scored), 35S:SPT (Met2) (nine plants), 35S:SPT (Met3) (nine plants), 35S:SPT (ΔAmphipatic) (five plants), 35S:SPT (ΔAcidic) (three plants), 35S:SPT (ΔNLS) (seven plants) and 35S:SPT (ΔC terminus) (eight plants), see Figure 5]. Bars extend for the SEM in each direction.

The next construct [35S:SPT (Met3)] showed that SPT protein deleted upstream of the third methionine (codon 107) could complement spt-2 defects to an intermediate level (Figure 6). The phenotype observed was similar to that seen in spt-1 plants (Figure 4), supporting the proposal that translation is reinitiated at this codon in the chain terminating spt-1 mutant (Heisler et al., 2001).

To test whether the conserved amphipathic helix and the acidic domains are required for SPT function, they were deleted individually in two constructs and inserted into spt-2 mutant plants (Figure 5b). The 35S:SPT (ΔAmphipathic) construct, lacking codons 47–57, inclusive, retained some ability to complement the mutant phenotype, but at a relatively low level (Figure 6). This is consistent with results from the 35S:SPT (Met3) deletion construct above, in which the amphipathic helix is also absent. (The slightly weaker complementation seen in the 35S:SPT ΔAmphipathic lines may be associated with differences in protein folding.) On the other hand, the 35S:SPT (ΔAcidic) construct, with codons 150–174 deleted, did not restore the phenotype significantly in the wild-type direction (Figure 6). The phenotype of all nine lines obtained was indistinguishable from the spt-2 mutant, indicating that this domain is absolutely required for SPT function.

As in other bHLH proteins (Heim et al., 2003; Toledo-Ortiz et al., 2003), SPT carries a conserved bipartite nuclear localization sequence (NLS) that overlaps with the N-terminal end of the bHLH domain. This region is apparently essential for SPT function, as when it was deleted in 35S:SPT (ΔNLS), a construct lacking codons 194–210, no complementation of the spt-2 mutant phenotype was seen in 21 independent transformed lines (Figure 6). To test the ability of this sequence to promote protein transport into the nucleus, the GFP sequence was translationally fused to the C terminus of the same SPT deletion [35S:SPT-GFP (ΔNLS)]. Compared with GFP fusions to the full-length SPT protein (35S:SPT-GFP), preferential accumulation in the nucleus was not seen (Figure 3c,d).

Finally, the consequence of removing the non-conserved C terminus of SPT downstream of the beta strand was tested by deleting the last 90 amino acids [35S:SPT (ΔC terminus)]. In the eight lines scored out of 15 independent insertion lines, the length, seed set and septum development were all restored close to the level seen in 35S:SPT controls (Figure 6).

It may be that the presence of a full-length but defective protein predicted to occur in spt-2 mutant plants has influenced the results, so similar tests were carried out by transforming spt-3 mutants with some of the constructs (Table S1). The spt-3 gene is predicted to encode a truncated protein lacking the bHLH region (Figure 5a). In each case, the spt-3 results were similar to those obtained with transformed spt-2 plants.

SPATULA may activate rather than repress its targets

Domains that artificially repress or activate target gene expression were added to SPT, and the developmental consequences in transformed wild-type plants were determined. The 35S:SPT construct was used, as it has no detectable effect on morphogenesis in transgenic wild-type plants (except for elongated hypocotyls; Penfield et al., 2005).

The SRDX transcriptional repression domain (Hiratsu et al., 2003) was translationally fused to the C terminus of SPT. All 17 independent transformants displayed defects in gynoecium development (Figure 7a–g). In 10 lines the gynoecium appeared normal externally, but internally the septum was defective (absent in five lines, partly absent in five lines; Figure 7f,g). In another five lines the style and upper region of the ovary were unfused: there were no stigmatic papillae and no seed set. They resembled the strongly affected gynoecia of spt mutant plants (Figure 7c,d). The gynoecial development in the last two lines were even more abnormal. Carpel fusion was disrupted further, and the apical regions extended into curved outgrowths (Figure 7e). This even stronger disruption may have occurred if redundantly regulated targets of SPT had then been inactivated. No other defects were seen, except in the most severely disrupted line that was dwarfed.

Figure 7.

 Consequences of artificially converting SPT into a transcriptional repressor or a transcriptional activator.
(a–g) Effect of the repressor fusion 35S:SPT-SRDX. (a–c) Medial views of gynoecia at stage 13 from a Landsberg erecta wild type (a), an spt-2 mutant (b) and a 35S:SPT-SRDX mutant (c), showing that the repressor causes similar disruptions to those seen in spt-2 mutants, with less stigmata and reduced apical fusion of the carpels. (d–g) Siliques of 35S:SPT-SRDX plants in Columbia showing even stronger apical disruptions to silique development (d, e), or weaker disruption externally (f) but full loss of the internal septum in the same silique (g).
(h–n) Effect of the activator fusion 35S:SPT-VP16.
(h, i) Vertical views of the inflorescence apex of wild type (h) and 35S:SPT-VP16 (i) plants, revealing that the latter has narrower, non-overlapping sepals, narrower petals and reduced anthers.
(j) SEM of a sepal showing stigmatic outgrowth at the apex, with smooth, style-like cells below and also scattered within other normal sepal epidermal cells (inset).
(k) SEM of the apex of a mature silique (medial view), showing lumpy valves and the presence of a pseudoreplum along the lateral edges (arrow).
(l,m) Siliques showing that growth disruptions are more severe in apical regions: (l) medial view of Columbia transformant and (m) the maximum extent of the pseudoreplum (lateral view of Landsberg transformant).
(n) Silique of a 35S:STYLISH1 transformant (lateral view) showing a pseudoreplum similar to that in (m).
(o–t) Ectopic expression of STYLISH2 in 35S:SPT-VP16 activated plants.
(o,p) STY2:GUS expression in an inflorescence (o) and mature flower (p) of the wild type.
(q–s) STY2:GUS expression in an inflorescence (q), mature flower (r) and silique (s) of plants carrying 35S:SPT-VP16, showing ectopic patches of expression in the sepals of young buds (arrow), and apical regions of the gynoecium and silique [compare with (l)].

The spatula-like phenotype in 35S:SPT-SRDX plants may have arisen through co-suppression. However, this is unlikely because RT-PCR tests showed that inflorescences of a representative sample of five T1 transformants contained normal and transgenic SPT mRNA, and, for the former, levels were not markedly less than those in untransformed plants. We conclude that it is unlikely that SPT normally represses its target genes.

To assess the consequences of the artificial activation of SPT targets, the VP16 activation domain (Triezenberg et al., 1988) was translationally fused to the C terminus of SPT, again under the control of the 35S promoter. Fifteen out of 22 independent transformants showed moderate growth disruption, with plants being smaller and slower growing than controls. Flower development was also disrupted to some extent (Figure 7h,i). Sepals were shorter and involuted, and they did not enclose the buds. In some cases, they showed carpelloid properties, especially when raised at higher temperatures (28°C). These included the presence of stigmatic papillae at the apex, immature style-like cells adjacent to the stigma and sometimes embedded in more basal epidermal cells (Figure 7j), and growth projections along the sepal edges that resemble ovule primordia. Petals had a narrower claw and wider blade than normal, and anthers were tapered at the apex and often did not release pollen (Figure 7i). The apical region of the siliques often displayed irregular and reduced valve development, evident by the constrained growth of the replum, and the lumpy appearance partly caused by the protruding seeds within (Figure 7k,l). Further irregularities included regions of abnormal thickening of the valve wall, and in some cases it appeared that the apical regions of each valve (along the lateral plane of the flower) had separated in two, with the regions between differentiating into a replum-like structure (Figure 7k,m). In this regard, they resembled the irregular gynoecium sometimes produced in plants overexpressing STY1 (Figure 7n). Thus, ectopic expression of SPT is unable to induce growth abnormalities, unless a strong activation domain is attached.

We tested the ability of the SPT protein to activate target reporter genes in yeast when bound to their promoter through the GAL4 DNA binding domain. The mean expression of a MEL1 reporter gene was not significantly different between 10 pGALDB-SPT transformants (1.24 ± 0.56 mU ml−1 X cell) and 10 negative control transformants (0.66 ± 0.26), whereas the result from eight full-length GAL4 positive controls was 42.84 ± 6.69 (all results are means ± SEM). This indicates that SPT does not act as an activator alone, although it is also consistent with a requirement for co-activators not found in yeast cells.

Ectopic SPT activates the STYLISH2 gene

STY1 and STY2 are two RING-finger genes with redundant functions in style development, and that genetically interact with SPT (Kuusk et al., 2002). Some of the abnormalities seen in the gynoecium of 35S:SPT-VP16 plants resemble defects observed when either STY1 or STY2 are constitutively overexpressed (Kuusk et al., 2002; Figure 7m,n). To test whether 35S:SPT-VP16 ectopically activates the STY genes, we compared the expression of STY2:GUS in 35S:SPT-VP16 plants (Figure 7q–s) with wild-type controls (Figure 7o,p). Strikingly, ectopic GUS expression was now visible in patches through the upper regions of abnormal gynoecia (Figure 7r,s). Similar patches were visible in carpelloid sepals (Figure 7q). This expression required the VP16 activation domain, as it was not seen in control 35S:SPT plants (not shown). To test whether STY2 expression normally requires SPT function, we examined its expression in spt-2 mutants. STY2 expression still clearly occurred (Figure 7t), so if it is a direct target of SPT regulation, then other transcription factors can also activate its expression.

The VP16 activation domain can compensate somewhat for the loss of the acidic domain

Finally, we tested whether the N-terminal region of the SPT protein, including the conserved amphipathic helix and acidic domains, is involved in the activation of target genes, by examining whether the VP16 activation domain could compensate for their loss. Another deletion construct, 35S:SPT (ΔN terminus), was made that removed all codons up to and including 187 (10 codons upstream of the bHLH domain) (Figure 5c). As expected, this had no effect when transformed into wild-type plants. However, when the activation domain was added [35S:SPT-VP16 (ΔN terminus)], it induced defects in transformed wild-type plants similar to those seen with the undeleted 35S:SPT-VP16 version. Sepal overlap was reduced in buds, petals were distorted and pollen was rarely shed from misshapen anthers. However, most abnormalities were considerably weaker, and gynoecia in the 35S:SPT-VP16 (ΔN terminus) lines were indistinguishable from normal.

To examine the possible gynoecium disruptions further, two deletion constructs, with and without VP16, were transformed into spt-2 and spt-3 mutant plants to test if they could complement the gynoecium defects. The construct without the VP16 domain [35S:SPT (ΔN terminus)] could not (15 spt-2 and 13 spt-3 lines examined), but significant restoration towards the wild type was seen in 35S:SPT-VP16 (ΔN terminus) lines, especially in the later formed flowers. For example, in seven of the 58 transformed spt-2 lines obtained, the septum of the first formed flower appeared on average along about half of the wild-type length (mean 46 ± 8%, range 8–71%), compared with its consistent absence in 10 spt-2 control plants. Out of 27 transformed spt-3 lines, 13 showed a full restoration of the septum in the tenth flower, whereas in 10 spt-3 controls the maximum fusion seen in the tenth flower was only 25%.

Thus the VP16 activation domain is able to compensate to some extent for the loss of the N-terminal region of the SPT protein that includes the conserved amphipathic helix and acidic domain.

Discussion

We have identified two new conserved structural domains within the SPT protein. One of these, the acidic domain, is apparently required for the carpel development function of SPT, whereas the other, the amphipathic helix, is not necessary for this function, but nevertheless supports it. Furthermore, these are the only two regions with a clearly predicted secondary structure conserved between the Arabidopsis and tomato orthologues of SPT (apart from the extended bHLH region), and yet the tomato protein can provide full SPT gynoecium function in Arabidopsis.

SPT activates its target genes in gynoecium development in association with co-activators

The present study provides evidence that SPT activates rather than represses its target carpel genes. The addition of an SRDX repression domain to the SPT protein phenocopies the consequences of the loss of SPT function in transgenic plants, whereas the addition of a VP16 activation domain can help restore SPT function in spt mutant plants. Recent evidence reveals that the PIF1 (PIL5) protein, a relative of SPT that interacts with phytochrome, directly activates target gene expression (Oh et al., 2007).

It also seems that SPT requires one or more co-activators to function appropriately, and that these are normally present only in regions where SPT is normally active. Ectopic expression of the SPT protein has consequences on development beyond the locations disrupted in spt mutant plants only when it is fused with the VP16 activation domain.

This 35S:SPT-VP16 construct is able to activate several components of the carpel development program, in that ectopic style-like tissues, stigmata and ovule-like outgrowths are present on the now involuted sepals. These abnormal sepals are similar to the first whorl organs of ap2 ag crc triple mutants, where SPT is known to be ectopically active (Alvarez and Smyth, 1999; Heisler et al., 2001). It seems that AP2 normally prevents SPT function, and presumably that of the proposed co-activators, in the first whorl. However, the lack of these co-activators can apparently be compensated for to some extent by the addition of the VP16 activator to SPT.

It is of interest that one of the genes ectopically activated by 35S:SPT-VP16 is STYLISH2 (Kuusk et al., 2002). However, it seems that STY2 is not a direct target of SPT activation, or that other transcription factors redundantly regulate its expression, as we have found that STY2 expression still occurs in stylar tissue when SPT function is compromised. Also, there is genetic evidence that the STY genes and SPT have overlapping rather than sequential style development functions (Kuusk et al., 2002).

STY and SPT gene functions are also linked through the involvement of auxin. Auxin plays a role in patterning the apical-to-basal polarity of the gynoecium (Nemhauser et al., 2000), with mutations of spt being rescued to some extent by the addition of inhibitors of polar auxin transport (Nemhauser et al., 2000), whereas sty1 mutant defects show enhanced susceptibility (Sohlberg et al., 2006). SPT may support the auxin signal transduction pathway in the apices of developing gynoecia (Nemhauser et al., 2000), whereas STY proteins may be involved in the control of auxin biosynthesis in apical regions (Sohlberg et al., 2006).

Other plant bHLH proteins have been shown to function in association with a co-activator that is a WD40 protein (Ramsay and Glover, 2005). The bHLH-WD40 combination interacts further with a range of different MYB transcription factors (through the bHLH unit) to regulate target genes involved in anthocyanin biosynthesis, and in epidermal cell differentiation. The sites of interaction of the bHLH component have not yet been associated with specific structural domains.

Possible roles of the acidic domain, the amphipathic helix and the beta strand

The acidic domain occurs as a predicted helical structure, with charged aspartate and glutamate residues localized on one surface. It is apparently associated with transcriptional activation, as the artificial activation sequence VP16 can compensate to some extent for its loss [assuming that this largely accounts for the inactivity of 35S:SPT (ΔN terminus)]. It may be that the acidic domain is the site of interaction of SPT with the proposed co-activators of transcription. Such an indirect role is more likely than direct activation of the basal transcription machinery. In some transcription factors, acidic regions have been identified as direct activators, but these are enriched for acidic residues, rather than having a strictly conserved sequence or structure (Ptashne and Gann, 1997). Amphipathic helices are also frequently the site of protein–protein interactions, and the conserved amphipathic helix may also interact with the proposed co-activator proteins.

The beta strand extension that lies adjacent to the bHLH region on its C terminal side is apparently important for SPT carpel function. A frame-shift mutation within it, spt-8, displays a strong phenotype, but a deletion commencing just downstream can almost fully complement the spt mutant phenotype. The amphipathic nature of this sequence is consistent with dimerization along its polar surface. It lies just downstream of the two helices of the bHLH region that are consistently involved in dimerization, so it could provide additional dimerization contact points. In this way its role would parallel that of the leucine zipper region in bHLH-Zip proteins (Blackwood and Eisenman, 1991), perhaps by promoting the specificity of its potential dimeric partners (Soucek et al., 1998). However, the only bHLH proteins reported to dimerize with SPT so far, the HECATE proteins (Gremski et al., 2007), do not carry this beta strand. They, too, are associated with septum and stigma development, and further study of their potential interaction with SPT will be of interest.

SPT has functions in addition to gynoecium and fruit development

Recently, SPT has been shown to be involved in regulating the germination of seeds in response to cold exposure, as well as in the expansion of cotyledons and petals (Penfield et al., 2005). In germinating seeds, SPT represses the production of the GA biosynthetic enzymes GA3ox1 and GA3ox2. If this repression is direct, it suggests that the SPT protein can act either as an activator (shown here in carpels and fruits) or as a repressor (in seeds). It is also of interest that the spt-2 mutant was found to act as a dominant negative disruption in germinating seeds (it is semi-dominant), but as a recessive mutation in its effects on cotyledon and petal expansion (Penfield et al., 2005). The latter is consistent with what we have observed for carpel and fruit development. The spt-2 protein presumably cannot bind DNA, and it may be that it is capable of interfering with normal SPT protein function in heterozygotes, but only in seeds, perhaps because the threshold of activity required for full SPT function is higher in these tissues.

The function of SPT has diverged from that of related bHLH proteins

The amphipathic helix and acidic domain cannot be identified in the other 14 members of the SPT subclade of bHLH proteins in Arabidopsis (group VII of Heim et al., 2003; subfamily 15 of Toledo-Ortiz et al., 2003). The only exception is ALACTRAZ where an amphipathic helix occupies codons 17–27 (Figure S1). ALC is the closest relative of SPT, and in the Arabidopsis fruit functions to define the separation layer of the dehiscence zone (Rajani and Sundaresan, 2001). SPT is also expressed in the dehiscence zone (Heisler et al., 2001), but there is no evidence that it plays a functional role there. Conversely, there is no evidence that ALC functions earlier in gynoecium development when SPT is active. Further study may uncover overlapping (redundant) functions of these two genes.

Among the other 13 members of this bHLH subclade, the only other function identified to date is the perception of light through phytochromes. Six bHLH proteins are generally known as PHYTOCHROME INTERACTING FACTOR (PIF) or PIF3-LIKE (PIL), and five have been shown to interact directly with phyB (Khanna et al., 2004). The domain necessary for this interaction, the active phytochrome binding (ABP) domain, occurs near the N terminus of the bHLH proteins, and is found in all members of this subclade except for SPT and ALC (Figure S1), and a divergent protein LONG HYPOCOTYL IN FAR RED1 (HFR1). As expected, SPT does not bind phyB (Khanna et al., 2004).

The divergence in SPT function from the other light-responding members of this subclade may have been relatively recent. Orthologues of SPT are clearly identifiable in a monocot (rice, Li et al., 2006) and in a gymnosperm P. glauca. However, these do not carry the amphipathic helix or the acidic domain, whereas all the eudicot sequences we examined do. Thus, the functions of these now conserved regions of the SPT protein, presumably associated with recruitment of new interacting partners and target genes associated with carpel and fruit development, may have arisen around the time of the emergence of the eudicots.

Experimental procedures

Plant strains

All spt mutants were isolated in the Landsberg erecta using EMS mutagenesis. The spt-1, spt-2 and spt-3 mutants have already been described (Alvarez and Smyth, 1999; Alvarez and Smyth 2002). The new spt-8 mutant was contributed by Yuval Eshed and John Bowman. Seeds carrying 35S:STY1 and STY2:GUS in the Columbia background were kindly provided by Eva Sundberg (Kuusk et al., 2002).

Generation of expression constructs

All 35S:SPT constructs were based on the SPT clone cDNA9 in pBluescript (Heisler et al., 2001). Truncated and deleted versions were generated by PCR. Truncations carried new methionine or stop codons, as appropriate. Deleted versions contained a SalI bridge across the deleted region that encoded valine and aspartate. These products were then inserted into the multiple cloning site of pART7 that contains a 5′ 35S promoter element and a 3′ OCS sequence, flanked by NotI restriction sites (Gleave, 1992). The NotI flanked inserts were then transferred to the NotI site of pMLBART, a derivative of pART27 (Gleave, 1992) in which the kanamycin plant resistance gene has been replaced by a glufosinate resistance gene. They were then transferred into Agrobacterium tumefaciens strain AGL1.

The tomato clone 35S:SlSPT was made using the expressed sequence tag (EST) cLEI11G2 (gi:7411210) obtained from the Clemson University Genomics Institute. A KpnI site was generated by PCR immediately upstream of the first AUG codon, and a BamHI site was generated immediately downstream of the cDNA sequence. This was inserted into pART7 and then pMLBART, as before.

The GFP gene was amplified from pBIN 35S:mgfp5-ER (Haseloff et al., 1996) (after removing the C-terminal HDEL sequence and creating a new stop codon), and was combined with the C terminus of SPT through an EcoRI bridge encoding glutamate and phenylalanine. The construct was called 35S:SPT-GFP. The NLS region (codons 194–210) was deleted by targetted PCR, substituting it with a SalI bridge encoding valine and alanine. The GFP sequence was attached to the C terminus of SPT in this construct, with a BamHI bridge encoding glycine and serine.

The SRDX repression domain of 12 codons (Hiratsu et al., 2003) was generated by PCR. The VP16 activation domain was amplified from the construct pFP14 (provided by John Bowman). In each case, the amplified products were fused to the C-terminal end of SPT using an EcoRI bridge in the same way as for the 35S:SPT-GFP construct above.

Plant transformation and selection of transgenic lines

Arabidopsis plants were transformed by infection with A. tumefaciens (Clough and Bent, 1998). In each transformation, approximately 10–30 independent transformants were obtained (Table S1). In many experiments, primary transformants (T1) were analysed, although in some experiments the most strongly complemented (or disrupted) T2 progeny plant of self-fertilized T1 plants was scored instead. For all insertion lines scored, the presence of the transgene was confirmed by PCR. In some cases, its expression was tested by RT-PCR. RNA was extracted from plants using the Qiagen RNeasy Plant Mini Kit, and RT-PCR was carried out using the Qiagen One Step RT-PCR kit. Primers spanned introns to distinguish transcripts from genes, with one of the primers usually present in the 3′-OCS sequence in pMLBART. As a control, expression of the ACTIN2 or APETALA1 gene was used. RT-PCR was also used to determine the product of the spt-8 mutant using primers within the SPT gene.

Reporter gene analysis

Staining to reveal expression of the STY2:GUS reporter gene, and assessment of nuclear localization in onion epidermal cells that had been biolistically transfected with GFP reporter constructs, followed the methodology of Brewer et al. (2004).

Yeast transcriptional activation assay

The BD Clontech Matchmaker kit (http://www.bd.com) was used. The full-length SPT cDNA9 was inserted into the plasmid pGBKT7 using artificial EcoRI and BamHI flanking sites, generated by PCR, to generate the prey construct pGAL4DB-SPT. This was transformed into yeast strain AH109, and transformants were selected by tryptophan prototrophy. They were tested for activation of HIS3 and ADE2 reporter genes, and assayed spectrophotmetrically in duplicate for the expression of the MEL1α-galactosidase reporter. The positive control was the full-length GAL4 plasmid pCL1, and the negative control was pGBKT7-Lam.

Scoring of siliques

Dried siliques on the main inflorescence shoot were scored, commencing from the first formed silique. Seed numbers were either counted directly or counted based on their funiculi, which are distinguishable for those of unfertilized ovules. To measure silique and septum length, siliques were dry mounted on microscope slides, and were matched against a 0.5-mm eyepiece grid using a dissection microscope. Silique length was estimated from the former point of attachment of the valves at the base of the ovary to the top of the dried stigma. Gaps in the septum mostly occurred at the apex of the silique, and the extent of the septum remaining was compared with its full potential length using the 0.5-mm grid.

Acknowledgements

We thank John Alvarez, Marcus Heisler and colleagues for advice and assistance, Yuval Eshed and John Bowman for spt-8, Eva Sundberg for 35S:STY1 and STY2:GUS seeds, Ruth N. Kaplan-Levy for the modified 35S:GFP construct, Annette Becker for unpublished EcSPT sequence information, and Martin O’Brien for carrying out several control PCR and RT-PCR assays. This work was supported in part by Australian Research Council grant A19927094.

Ancillary