Mechanistic insight into a peptide hormone signaling complex mediating floral organ abscission

1During their growth, development and reproduction plants use cell separation processes to detach no-longer required, damaged or senescent organs. Abscission of floral organs in Arabidopsis is a model system to study these cell separation processes in molecular detail. The LRR-RKs HAESA (greek: to adhere to) and HAESA-LIKE 2 (HSL2) redundantly control floral abscission. Loss-of-function of the secreted small protein INFLORESCENCE DEFICIENT IN ABSCISSION (IDA) causes floral organs to remain attached while its over-expression leads to premature shedding. Full-length IDA is proteolytically processed and a conserved stretch of 20 amino-acids (termed EPIP) can rescue the IDA loss-of-function phenotype (Figure 1A). It has been demonstrated that a dodecamer peptide within EPIP is able to activate HAESA and HSL2 in transient assays in tobacco cells. This sequence motif is highly conserved among IDA family members (IDA-LIKE PROTEINS, IDLs) and contains a central Pro residue, presumed to be post-translationally modified to hydroxyproline (Hyp; Figure 1A). The available genetic and biochemical evidence suggests that IDA and HAESA together control floral abscission, but it is poorly understood if IDA is directly sensed by the receptor kinase HAESA and how IDA binding at the cell surface would activate the receptor.

2We purified the HAESA ectodomain (residues 20–620) from baculovirus-infected insect cells (Figure 1—figure supplement 1A, see Materials and methods) and quantified the interaction of the ~75 kDa glycoprotein with synthetic IDA peptides using isothermal titration calorimetry (ITC). A Hyp-modified dodecamer comprising the highly conserved PIP motif in IDA (Figure 1A) interacts with HAESA with 1:1 stoichiometry (N) and with a dissociation constant (Kd) of ~20 μM (Figure 1B). We next determined crystal structures of the apo HAESA ectodomain and of a HAESA-IDA complex, at 1.74 and 1.86 Å resolution, respectively (Figure 1C; Figure 1—figure supplement 1B–D; Tables 1,2). IDA binds in a completely extended conformation along the inner surface of the HAESA ectodomain, covering LRRs 2–14 (Figure 1C,D, Figure 1—figure supplement 2). The central Hyp64IDA is buried in a specific pocket formed by HAESA LRRs 8–10, with its hydroxyl group establishing hydrogen bonds with the strictly conserved Glu266HAESA and with a water molecule, which in turn is coordinated by the main chain oxygens of Phe289HAESA and Ser311HAESA (Figure 1E; Figure 1—figure supplement 3). The restricted size of the Hyp pocket suggests that IDA does not require arabinosylation of Hyp64IDA for activity in vivo, a modification that has been reported for Hyp residues in plant CLE peptide hormones. The C-terminal Arg-His-Asn motif in IDA maps to a cavity formed by HAESA LRRs 11–14 (Figure 1D,F). The COO- group of Asn69IDA is in direct contact with Arg407HAESA and Arg409HAESA and HAESA cannot bind a C-terminally extended IDA-SFVN peptide (Figures 1D,F, 2D). This suggests that the conserved Asn69IDA may constitute the very C-terminus of the mature IDA peptide in planta and that active IDA is generated by proteolytic processing from a longer pre-protein. Mutation of Arg417HSL2 (which corresponds to Arg409HAESA) causes a loss-of-function phenotype in HSL2, which indicates that the peptide binding pockets in different HAESA receptors have common structural and sequence features. Indeed, we find many of the residues contributing to the formation of the IDA binding surface in HAESA to be conserved in HSL2 and in other HAESA-type receptors in different plant species (Figure 1—figure supplement 3). A N-terminal Pro-rich motif in IDA makes contacts with LRRs 2–6 of the receptor (Figure 1D, Figure 1—figure supplement 2A–C). Other hydrophobic and polar interactions are mediated by Ser62IDA, Ser65IDA and by backbone atoms along the IDA peptide (Figure 1D, Figure 1—figure supplement 2A–C).

3We next investigated whether HAESA binds N-terminally extended versions of IDA. We obtained a structure of HAESA in complex with a PKGV-IDA peptide at 1.94 Å resolution (Table 2). In this structure, no additional electron density accounts for the PKGV motif at the IDA N-terminus (Figure 2A,B). Consistently, PKGV-IDA and IDA have similar binding affinities in our ITC assays, further indicating that HAESA senses a dodecamer peptide comprising residues 58-69IDA (Figure 2D).

4We next tested if HAESA binds other IDA peptide family members. IDL1, which can rescue IDA loss-of-function mutants when introduced in abscission zone cells, can also be sensed by HAESA, albeit with lower affinity (Figure 2D). A 2.56 Å co-crystal structure with IDL1 reveals that different IDA family members use a common binding mode to interact with HAESA-type receptors (Figure 2A–C,E, Table 2). We do not detect interaction between HAESA and a synthetic peptide missing the C-terminal Asn69IDA (ΔN69), highlighting the importance of the polar interactions between the IDA carboxy-terminus and Arg407HAESA/Arg409HAESA (Figures 1F, 2D). Replacing Hyp64IDA, which is common to all IDLs, with proline impairs the interaction with the receptor, as does the Lys66IDA/Arg67IDA → Ala double-mutant discussed below (Figure 1A, 2D). Notably, HAESA can discriminate between IDLs and functionally unrelated dodecamer peptides with Hyp modifications, such as CLV3 (Figures 2D, 7).

5Our binding assays reveal that IDA family peptides are sensed by the isolated HAESA ectodomain with relatively weak binding affinities (Figures 1B, 2A–D). It has been recently reported that SOMATIC EMBRYOGENESIS RECEPTOR KINASES (SERKs) are positive regulators of floral abscission and can interact with HAESA and HSL2 in an IDA-dependent manner. As all five SERK family members appear to be expressed in the Arabidopsis abscission zone, we quantified their relative contribution to floral abscission in Arabidopsis using a petal break-strength assay. Our experiments suggest that among the SERK family members, SERK1 is a positive regulator of floral abscission. We found that the force required to remove the petals of serk1-1 mutants is significantly higher than that needed for wild-type plants, as previously observed for haesa/hsl2 mutants, and that floral abscission is delayed in serk1-1 (Figure 3A). The serk2-2, serk3-1, serk4-1 and serk5-1 mutant lines showed a petal break-strength profile not significantly different from wild-type plants. Possibly because SERKs have additional roles in plant development such as in pollen formation and brassinosteroid signaling, we found that higher-order SERK mutants exhibit pleiotropic phenotypes in the flower, rendering their analysis and comparison by quantitative petal break-strength assays difficult. We thus focused on analyzing the contribution of SERK1 to HAESA ligand sensing and receptor activation.

6In vitro, the LRR ectodomain of SERK1 (residues 24–213) forms stable, IDA-dependent heterodimeric complexes with HAESA in size exclusion chromatography experiments (Figure 3B). We next quantified the contribution of SERK1 to IDA recognition by HAESA. We found that HAESA senses IDA with a ~60 fold higher binding affinity in the presence of SERK1, suggesting that SERK1 is involved in the specific recognition of the peptide hormone (Figure 3C). We next titrated SERK1 into a solution containing only the HAESA ectodomain. In this case, there was no detectable interaction between receptor and co-receptor, while in the presence of IDA, SERK1 strongly binds HAESA with a dissociation constant in the mid-nanomolar range (Figure 3C). This suggests that IDA itself promotes receptor – co-receptor association, as previously described for the steroid hormone brassinolide and for other LRR-RK complexes. Importantly, hydroxyprolination of IDA is critical for HAESA-IDA-SERK1 complex formation (Figure 3C,D). Our calorimetry experiments now reveal that SERKs may render HAESA, and potentially other receptor kinases, competent for high-affinity sensing of their cognate ligands.

7Upon IDA binding at the cell surface, the kinase domains of HAESA and SERK1, which have been shown to be active protein kinases, may interact in the cytoplasm to activate each other. Consistently, the HAESA kinase domain can transphosphorylate SERK1 and vice versa in in vitro transphosphorylation assays (Figure 3E). Together, our genetic and biochemical experiments implicate SERK1 as a HAESA co-receptor in the Arabidopsis abscission zone.

8To understand in molecular terms how SERK1 contributes to high-affinity IDA recognition, we solved a 2.43 Å crystal structure of the ternary HAESA – IDA – SERK1 complex (Figure 4A, Table 2). HAESA LRRs 16–21 and its C-terminal capping domain undergo a conformational change upon SERK1 binding (Figure 4B). The SERK1 ectodomain interacts with the IDA peptide binding site using a loop region (residues 51-59SERK1) from its N-terminal cap (Figure 4A,C). SERK1 loop residues establish multiple hydrophobic and polar contacts with Lys66IDA and the C-terminal Arg-His-Asn motif in IDA (Figure 4C). SERK1 LRRs 1–5 and its C-terminal capping domain form an additional zipper-like interface with residues originating from HAESA LRRs 15–21 and from the HAESA C-terminal cap (Figure 4D). SERK1 binds HAESA using these two distinct interaction surfaces (Figure 1—figure supplement 3), with the N-cap of the SERK1 LRR domain partially covering the IDA peptide binding cleft.

9The four C-terminal residues in IDA (Lys66IDA-Asn69IDA) are conserved among IDA family members and are in direct contact with SERK1 (Figures 1A, 4C). We thus assessed their contribution to HAESA – SERK1 complex formation. Deletion of the buried Asn69IDA completely inhibits receptor – co-receptor complex formation and HSL2 activation (Figure 5A,B). A synthetic Lys66IDA/Arg67IDA → Ala mutant peptide (IDA K66A/R66A) showed a 10 fold reduced binding affinity when titrated in a HAESA/SERK1 protein solution (Figures 5A,B, 2D). We over-expressed full-length wild-type IDA or this Lys66IDA/Arg67IDA → Ala double-mutant to similar levels in Col-0 Arabidopsis plants (Figure 5D). We found that over-expression of wild-type IDA leads to early floral abscission and an enlargement of the abscission zone (Figure 5C–E). In contrast, over-expression of the IDA Lys66IDA/Arg67IDA → Ala double mutant significantly delays floral abscission when compared to wild-type control plants, suggesting that the mutant IDA peptide has reduced activity in planta (Figure 5C–E). Comparison of 35S::IDA wild-type and mutant plants further indicates that mutation of Lys66IDA/Arg67IDA → Ala may cause a weak dominant negative effect (Figure 5C–E). In agreement with our structures and biochemical assays, this experiment suggests a role of the conserved IDA C-terminus in the control of floral abscission.

10In contrast to animal LRR receptors, plant LRR-RKs harbor spiral-shaped ectodomains and thus they require shape-complementary co-receptor proteins for receptor activation. For a rapidly growing number of plant signaling pathways, SERK proteins act as these essential co-receptors. SERK1 has been previously reported as a positive regulator in plant embryogenesis, male sporogenesis, brassinosteroid signaling and in phytosulfokine perception. Recent findings by and our mechanistic studies now also support a positive role for SERK1 in floral abscission. As serk1-1 mutant plants show intermediate abscission phenotypes when compared to haesa/hsl2 mutants, SERK1 likely acts redundantly with other SERKs in the abscission zone (Figure 3A). It has been previously suggested that SERK1 can inhibit cell separation. However our results show that SERK1 also can activate this process upon IDA sensing, indicating that SERKs may fulfill several different functions in the course of the abscission process.

11While the sequence of the mature IDA peptide has not been experimentally determined in planta, our HAESA-IDA complex structures and calorimetry assays suggest that active IDLs are hydroxyprolinated dodecamers. It will be thus interesting to see if proteolytic processing of full-length IDA in vivo is regulated in a cell-type or tissue-specific manner. The central Hyp residue in IDA is found buried in the HAESA peptide binding surface and thus this post-translational modification may regulate IDA bioactivity. Our comparative structural and biochemical analysis further suggests that IDLs share a common receptor binding mode, but may preferably bind to HAESA, HSL1 or HSL2 in different plant tissues and organs.

12In our quantitative biochemical assays, the presence of SERK1 dramatically increases the HAESA binding specificity and affinity for IDA. This observation is consistent with our complex structure in which receptor and co-receptor together form the IDA binding pocket. The fact that SERK1 specifically interacts with the very C-terminus of IDLs may allow for the rational design of peptide hormone antagonists, as previously demonstrated for the brassinosteroid pathway. Importantly, our calorimetry assays reveal that the SERK1 ectodomain binds HAESA with nanomolar affinity, but only in the presence of IDA (Figure 3C). This ligand-induced formation of a receptor – co-receptor complex may allow the HAESA and SERK1 kinase domains to efficiently trans-phosphorylate and activate each other in the cytoplasm. It is of note that our reported binding affinities for IDA and SERK1 have been measured using synthetic peptides and the isolated HAESA and SERK1 ectodomains, and thus might differ in the context of the full-length, membrane-embedded signaling complex.

13Comparison of our HAESA – IDA – SERK1 structure with the brassinosteroid receptor signaling complex, where SERK1 also acts as co-receptor, reveals an overall conserved mode of SERK1 binding, while the ligand binding pockets map to very different areas in the corresponding receptors (LRRs 2 – 14; HAESA; LRRs 21 – 25, BRI1) and may involve an island domain (BRI1) or not (HAESA) (Figure 6A). Several residues in the SERK1 N-terminal capping domain (Thr59SERK1, Phe61SERK1) and the LRR inner surface (Asp75SERK1, Tyr101SERK1, SER121SERK1, Phe145SERK1) contribute to the formation of both complexes (Figures 4C,D, 6B). In addition, residues 53-55SERK1 from the SERK1 N-terminal cap mediate specific interactions with the IDA peptide (Figures 4C, 6B). These residues are not involved in the sensing of the steroid hormone brassinolide. In both cases however, the co-receptor completes the hormone binding pocket. This fact together with the largely overlapping SERK1 binding surfaces in HAESA and BRI1 allows us to speculate that SERK1 may promote high-affinity peptide hormone and brassinosteroid sensing by simply slowing down dissociation of the ligand from its cognate receptor.

14Our experiments reveal that SERK1 recognizes a C-terminal Arg-His-Asn motif in IDA. Importantly, this motif can also be found in other peptide hormone families (Figure 7). Among these are the CLE peptides regulating stem cell maintenance in the shoot and the root. It is interesting to note, that CLEs in their mature form are also hydroxyprolinated dodecamers, which bind to a surface area in the BARELY ANY MERISTEM 1 receptor that would correspond to part of the IDA binding cleft in HAESA. Diverse plant peptide hormones may thus also bind their LRR-RK receptors in an extended conformation along the inner surface of the LRR domain and may also use small, shape-complementary co-receptors for high-affinity ligand binding and receptor activation.

15Synthetic genes coding for the Arabidopsis thaliana HAESA (residues 20–620) and SERK1 ectodomains (residues 24–213, carrying Asn115→Asp and Asn163→Gln mutations), codon optimized for expression in Trichoplusia ni (Geneart, Germany), were cloned into a modified pBAC-6 transfer vector (Novagen, Billerica, MA), providing an azurocidin signal peptide and a C-terminal TEV (tobacco etch virus protease) cleavable Strep-9xHis tandem affinity tag. Recombinant baculoviruses were generated by co-transfecting transfer vectors with linearised baculovirus DNA (ProFold-ER1, AB vector, San Diego, CA) followed by viral amplification in Spodoptera frugiperda Sf9 cells. The HAESA and SERK1 ectodomains were individually expressed in Trichoplusia ni Tnao38 cells using a multiplicity of infection of 3, and harvested from the medium 2 days post infection by tangential flow filtration using 30 kDa MWCO and 10 kDa MWCO (molecular weight cut-off) filter membranes (GE Healthcare Life Sciences, Pittsburgh, PA), respectively. Proteins were purified separately by sequential Ni2+ (HisTrap HP, GE Healthcare) and Strep (Strep-Tactin Superflow high-capacity, IBA, Germany) affinity chromatography. Next, affinity tags were removed by incubating the purified proteins with recombinant Strep-tagged TEV protease in 1:100 molar ratio. The cleaved tag and the protease were separated from HAESA and SERK1 by a second Strep affinity step. The purified HAESA ectodomain was incubated with a synthetic IDA peptide (YVPIPPSA-Hyp-SKRHN, the N-terminal Tyr residue was added to allow for peptide quantification by UV absorbance) and the SERK1 ectodomain in 1:1:1.5 molar ratio. The HAESA-IDA-SERK1 complex was purified by size exclusion chromatography on a Superdex 200 HR10/30 column (GE Healthcare) equilibrated in 20 mM citric acid pH 5.0, 100 mM NaCl). Peak fractions containing the complex were concentrated to ~10 mg/mL and immediately used for crystallization. About 0.2 mg of purified HAESA and 0.1 mg of purified SERK1 protein were obtained from 1 L of insect cell culture, respectively.

16Hexagonal crystals of the isolated HAESA ectodomain developed at room-temperature in hanging drops composed of 1.0 μL of protein solution (5.5 mg/mL) and 1.0 μL of crystallization buffer (21% PEG 3,350, 0.2 M MgCl2 · 6 H2O, 0.1 M citric acid pH 4.0), suspended above 1.0 mL of crystallization buffer. For structure solution crystals were derivatized and cryo-protected by serial transfer into crystallization buffer supplemented with 0.5 M NaI and 15% ethylene glycol and cryo-cooled in liquid nitrogen. Redundant single-wavelength anomalous diffraction (SAD) data to 2.39 Å resolution were collected at beam-line PXII at the Swiss Light Source (SLS), Villigen, CH with λ=1.7 Å. A native data set to 1.74 Å resolution was collected on a crystal from the same drop cryo-protected by serial transfer into crystallization buffer supplemented with 15% (v/v) ethylene glycol only (λ=1.0 Å; Table 1).

17HAESA complexes with IDA (PIPPSA-Hyp-SKRHN), PKGV-IDA (YPKGVPIPPSA-Hyp-SKRHN) and IDL1 (LVPPSG-Hyp-SMRHN) peptide hormones were obtained by soaking apo crystals in crystallization buffer containing the respective synthetic peptide at a final concentration of 15 mM. Soaked crystals diffracted to 1.86 Å (HAESA – IDA), 1.94 Å (HAESA-PKGV-IDA) and 2.56 Å resolution (HAESA – IDL1), respectively (Table 2). Orthorhombic crystals of the HAESA-IDA-SERK1 complex developed in 18% PEG 8000, MgCl2 · 6 H2O, 0.1 M citric acid and diffracted to 2.43 Å resolution (Table 2). Data processing and scaling was done in XDS (version: Nov 2014).

18The SAD method was used to determine the structure of the isolated HAESA ectodomain. SHELXD located 32 iodine sites (CC All/Weak 37.7/14.9). 20 consistent sites were input into the program SHARP for phasing and identification of 8 additional sites at 2.39 Å resolution. Refined heavy atom sites and phases were provided to PHENIX.AUTOBUILD for density modification and automated model building. The structure was completed in alternating cycles of model building in COOT and restrained TLS refinement in REFMAC5 (version 5.8.0107) against an isomorphous high resolution native data set. Crystals contain one HAESA monomer per asymmetric unit with a solvent content of ~55%, the final model comprises residues 20 – 615. The refined structure has excellent stereochemistry, with 93.8% of all residues in the favored region of the Ramachandran plot, no outliers and a PHENIX.MOLPROBITY score of 1.34 (Table 1).

19The HAESA – IDA – SERK1 complex structure was determined by molecular replacement with the program PHASER, using the isolated HAESA and SERK1 (PDB-ID: 4LSC) LRR domain structures as search models. The solution comprises one HASEA-IDA-SERK1 complex in the asymmetric unit. The structure was completed in iterative cycles of manual model-building in COOT and restrained TLS refinement in REFMAC5. Amino acids whose side-chain position could not be modeled with confidence were truncated to alanine (0.6 – 1% of total residues), the stereochemistry of N-linked glycan structures was assessed with the CCP4 program PRIVATEER-VALIDATE. The refined model has 94.44% of all residues in the favored region of the Ramachandran plot, no outliers and a PHENIX.MOLPROBITY score of 1.17 (Table 2). Structural visualization was done with POVScript+ and POV-Ray (http://www.povray.org).

20Gel filtration experiments were performed using a Superdex 200 HR 10/30 column (GE Healthcare) pre-equilibrated in 20 mM citric acid (pH 5) and 100 mM NaCl. 100 μL of the isolated HAESA ectodomain (5.5 mg/mL), of the purified SERK1 LRR domain (3 mg/mL) or of mixtures of HAESA and SERK1 (either in the presence or absence of synthetic wild-type IDA, wild-type IDL1 or mutant IDA peptides at a concentration of 25 μM; 10 mg/mL; samples contained HAESA and SERK1 in 1:1 molar ratio) were loaded sequentially onto the column and elution at 0.5 mL/min was monitored by ultraviolet absorbance at 280 nm.

21ITC experiments were performed using a Nano ITC (TA Instruments, New Castle, DE) with a 1.0 mL standard cell and a 250 μL titration syringe. Proteins were dialyzed extensively against ITC buffer (20 mM citric acid pH 5.0, 100 mM NaCl) and synthetic wild-type or point-mutant peptides (with wild-type IDA sequence YVPIPPSA-Hyp-SKRHN, PKGV-IDA YPKGVPIPPSA-Hyp-SKRHN, IDA-SFVN YPIPPSA-Hyp-SKRHNSFVN, IDL1 YLVPPSG-Hyp-SMRHN and CLV3 sequence YRTV-Hyp-SG-Hyp-DPLHH) were dissolved in ITC buffer prior to all titrations. Molar protein concentrations for SERK1 and HAESA were calculated using their molar extinction coefficient and a molecular weight of 27,551 and 74,896 Da, respectively (determined by MALDI-TOF mass spectrometry). Experiments were performed at 25°C. A typical experiment consisted of injecting 10 μL aliquots of peptide solution (250 μM) into 20 μM HAESA. The concentrations for the complex titrations were 150 μM of ligand (either wild-type or point-mutant IDA peptides) in the syringe and 10 μM of a 1:1 HAESA – SERK1 protein mixture in the cell at time intervals of 150 s to ensure that the titration peak returned to the baseline. Binding of SERK1 to HAESA was assessed by titrating SERK1 (100 μM) into a solution containing HAESA (10 μM) in the pre- or absence of 150 μM wild-type IDA peptide. ITC data were corrected for the heat of dilution by subtracting the mixing enthalpies for titrant solution injections into protein free ITC buffer. Data were analyzed using the NanoAnalyze program (version 2.3.6) as provided by the manufacturer.

22Coding sequences of SERK1 kinase domain (SERK1-KD) (residues 264–625) and HAESA-KD (residues 671–969) were cloned into a modified pET (Novagen) vector providing an TEV-cleavable N-terminal 8xHis-StrepII-Thioredoxin tag. Point mutations were introduced into the SERK1 (Asp447→Asn; mSERK1) and HAESA (Asp837→Asn; mHAESA) coding sequences by site directed mutagenesis, thereby rendering the kinases inactive. The plasmids were transformed into E.coli Rosetta 2 (DE3) (Novagen). Protein expression was induced by adding IPTG to final concentration of 0.5 mM to cell cultures grown to an OD600 = 0.6. Cells were then incubated at 16°C for 18 hr, pelleted by centrifugation at 5000 x g and 4°C for 15 min, and resuspended in buffer A (20 mM Tris-HCl pH 8, 500 mM NaCl, 4 mM MgCl2 and 2 mM β-Mercaptoethanol) supplemented with 15 mM Imidazole and 0.1% (v/v) Igepal. After cell lysis by sonication, cell debris was removed by centrifugation at 35,000 x g and 4°C for 30 min. The recombinant proteins were isolated by Co2+ metal affinity purification using a combination of batch and gravity flow approaches (HIS-Select Cobalt Affinity Gel, Sigma, St. Louis, MO). After washing the resin with the wash buffer (buffer A + 15 mM Imidazole) proteins were eluted in buffer A supplemented with 250 mM Imidazole. All elutions were then dialyzed against 20 mM Tris-HCl pH 8, 250 mM NaCl, 4 mM MgCl2 and 0.5 mM TCEP. For SERK1-KD and mSERK1-KD the 8xHis-StrepII-Thioredoxin tag was removed with 6xHis tagged TEV protease. TEV and the cleaved tag were removed by a second metal affinity purification step. Subsequently, all proteins were purified by gel filtration on a Superdex 200 10/300 GL column equilibrated in 20 mM Tris pH 8, 250 mM NaCl, 4 mM MgCl2 and 0.5 mM TCEP. Peak fractions were collected and concentrated using Amicon Ultra centrifugation devices (10,000 MWCO). For in vitro kinase assays, 1 μg of HAESA-KD, 0.25 μg of SERK1-KD and 2 μg of mSERK1 and mHAESA were used in a final reaction volume of 20 μl. The reaction buffer consisted of 20 mM Tris pH 8, 250 mM NaCl, 4 mM MgCl2 and 0.5 mM TCEP. The reactions were started by the addition of 4 μCi [γ-32P]-ATP (Perkin-Elmer, Waltham, MA), incubated at room temperature for 45 min and stopped by the addition of 6x SDS-loading dye immediately followed by incubating the samples at 95°C. Proteins of the whole reaction were subsequently separated via SDS-PAGE in 4–15% gradient gels (TGX, Biorad, Hercules, CA) and stained with Instant Blue (Expedeon, San Diego, CA). After pictures were taken of the stained gel, 32P-derived signals were visualized by exposing the gel to an X-ray film (Fuji, SuperRX, Valhalla, NY).

2335S::IDA wild-type and 35S::IDA (R66 → Ala/K67 → Ala) over-expressing transgenic lines in Col-0 background were generated as follows: The constructs were introduced in the destination vector pB7m34GW2 and transferred to A. tumefaciens strain pGV2260. Plants were transformed using the floral dip method. Transformants were selected in medium supplemented with BASTA up to the T3 generation. For phenotyping, plants were grown at 21°C with 50% humidity and a 16h light: 8 hr dark cycle.

24Plants were grown on ½ Murashige and Skoog (MS) plates supplemented with 1% sucrose. After 7 d, ∼30 to 40 seedlings were collected and frozen in liquid nitrogen. Total RNA was extracted using a RNeasy plant mini kit (Qiagen, Valencia, CA), and 1 μg of the RNA solution obtained was reverse-transcribed using the SuperScritpVILO cDNA synthesis kit (Invitrogen, Grand Island, NY). RT-qPCR amplifications and measurements were performed using a 7900HT Fast Real Time PCR-System by Applied Biosystems (Carlsbad, CA). RT-qPCR amplifications were monitored using SYBR-Green fluorescent stain (Applied Biosystems). Relative quantification of gene expression data was performed using the 2−ΔΔCT (or comparative CT) method. Expression levels were normalized using the CT values obtained for the actin2 gene (forward: TGCCAATCTACGAGGGTTTC; reverse: TTCTCGATGGAAGAGCTGGT). For detection and amplification of IDA sequence we used specific primers (forward: TCGTACGATGATGGTTCTGC; reverse: GAATGGGAACGCCTTTAGGT). The presence of a single PCR product was further verified by dissociation analysis in all amplifications. All quantifications were made in quadruplicates on RNA samples obtained from three independent experiments.

25serk1-1, serk2-2, serk3-1, serk4-1 and serk5-1 and Col-0 wild-type plants were grown in growth chambers at 22°C under long days (16 hr day/8 hr dark) at a light intensity of 100 µE·m-2·sec-1. Petal break-strength was quantified as the force in gram equivalents required for removal of a petal from a flower when the plants had a minimum of twenty flowers and siliques. Measurements were performed using a load transducer as described in. Break-strength was measured for 15 plants and a minimum of 15 measurements at each position.