The amino terminal F-box domain of Petunia inflata S-locus F-box protein is involved in the S-RNase-based self-incompatibility mechanism.

BACKGROUND AND AIMS: Pistils of flowering plants possessing self-incompatibility (SI) can distinguish between self and non-self pollen, and only allow non-self pollen to effect fertilization. For Petunia inflata, the S-RNase gene encodes pistil specificity and multiple S-locus F-box (SLF) genes encode pollen specificity. Each SLF produced in pollen interacts with a subset of non-self S-RNases to mediate their ubiquitination and degradation by the 26S proteasome. RATIONALE: S-locus F-box has been proposed to function as a component of the conventional SCF (SKP1-CULLIN-F-box protein) complex, based on the finding that two SKP1-like proteins, AhSSK1 (Antirrhinum hispanicum SLF-interacting SKP1-like1) and PhSSK1 (Petunia hybrida SSK1), interact with the F-box domain of some allelic variants of SLF. However, we previously showed that PiSLF (P. inflata SLF) did not interact with any SKP1 of P. inflata or Arabidopsis thaliana, but instead interacted with a RING-finger protein, PiSBP1 (P. inflata S-RNase-Binding Protein1), which may also play the role of SKP1. Thus, the biochemical nature of the SLF-containing complex is as yet unclear. PRINCIPAL
RESULTS: To examine whether the F-box domain of PiSLF is required for SI function, we expressed a truncated PiSLF(2) (S(2) allelic variant) without this domain in S(2)S(3) plants and showed that, unlike the full-length PiSLF(2), it did not cause breakdown of SI in S(3) pollen. We identified PiSSK1 (P. inflata SSK1) and found that it did not interact with PiSLF(1), PiSLF(2) or PiSLF(3).
CONCLUSIONS: The finding that the truncated PiSLF(2) did not cause breakdown of SI in S(3) transgenic pollen suggests that the F-box domain of PiSLF(2) is required for mediating degradation of S(3)-RNase, a non-self S-RNase, in S(3) pollen, and thus is required for SI function. The finding that PiSSK1 did not interact with three allelic variants of PiSLF is consistent with our previous finding that PiSLF might not be in a conventional SCF complex.

Introduction

Self-incompatibility (SI) is a reproductive strategy adopted by many flowering plants to reject self pollen but accept non-self pollen for fertilization (de Nettancourt 2001). Self-incompatibility is controlled by the highly polymorphic S-locus, which contains the genes for female and male specificity determinants in SI, respectively. Variants of the S-locus are termed S-haplotypes, and are designated as S, S, S, etc.

For the type of SI possessed by Solanaceae, Rosaceae and Plantaginaceae, pollen is recognized as self pollen if its S-haplotype matches either S-haplotype of the diploid pistil (Kao and Tsukamoto 2004; Franklin-Tong 2008). Both self and non-self pollen germinate on the stigmatic surface, but the growth of self pollen tubes is arrested typically in the upper third segment of the style. The female specificity determinant is encoded by the highly polymorphic S-RNase gene (Lee ; Murfett ), which produces a T2-type ribonuclease (McClure ). Through genomic sequencing of the S-locus, another polymorphic gene, S-locus F-box (SLF or SFB), was identified first in Antirrhinum (Plantaginaceae) (Lai ) and then in several rosaceous species (Entani ; Ushijima ; Yamane ) and Petunia inflata (Solanaceae) (Wang ). S-locus F-box was so named because its protein product contains a predicted F-box domain at the N-terminus. In P. inflata, PiSLF (S allele of P. inflata SLF; type-1 SLF) was identified ∼161 kb downstream from the S-RNase gene in the S-haplotype genome. The function of PiSLF in controlling pollen SI specificity was established via a transgenic approach (Sijacic ) designed based on an old observation that SI breaks down in diploid heteroallelic pollen carrying two different pollen S-alleles, but not in homoallelic pollen carrying two copies of the same S-allele (Brewbaker and Natarajan 1960). For example, when PiSLF was introduced into S transgenic plants, it caused breakdown of SI in S pollen (heteroallelic pollen carrying both S- and S-alleles of PiSLF) but not in S pollen (homoallelic pollen carrying two copies of S-allele of PiSLF) (Sijacic ).

S-RNase is synthesized in the transmitting cell of the style and secreted into the extracellular space of the transmitting tract. After germinating on the stigmatic surface and penetrating into the transmitting tract of the style, a pollen tube takes up both self and non-self S-RNases by an as yet unknown mechanism (Goldraij ). As predicted by a protein degradation model, PiSLF, PiSBP1 (P. inflata S-RNase-Binding Protein1; a RING-finger protein) and a CULLIN-1-like protein form a novel E3 ubiquitin ligase complex, which specifically targets any non-self S-RNases for ubiquitination and ultimate degradation by the 26S proteasome inside the pollen tube (Hua and Kao 2006; Hua ). Thus, in the case of incompatible pollination wherein self S-RNase is taken up by a pollen tube, the self S-RNase degrades RNA in the cytoplasm of the pollen tube to result in its growth arrest, and in the case of compatible pollination, the pollen tube uses its PiSLF-containing E3 complex to detoxify non-self S-RNases, allowing tube growth to effect fertilization (Hua ).

Very recently, through in vivo functional assay of additional alleles of SLF of P. inflata, Petunia hybrida and Petunia axillaris, it was discovered that the control of pollen specificity in Petunia is more complex than initially thought, as the pollen determinant is encoded by multiple types of polymorphic SLF genes, and not just the type of SLF gene first identified by sequencing of the S-locus (Kubo ). Co-immunoprecipitation experiments using extracts of transgenic pollen expressing an allele of a particular type of SLF and style extracts containing either self or non-self S-RNases (Kubo ) have further confirmed a previously discovered key biochemical feature of the protein degradation model that an SLF interacts more strongly with its non-self S-RNases than with its self S-RNase (Hua and Kao 2006). A modified protein degradation model, named collaborative non-self recognition, has been proposed. According to this model, for a given S-haplotype, each type of SLF can only recognize and interact with a subset of non-self S-RNases, and multiple types of SLF proteins are required to collaboratively recognize all non-self S-RNases to mediate their degradation to allow cross-compatible pollinations. However, none of the SLF proteins is able to efficiently bind self S-RNase to result in its degradation, thus allowing it to exert its cytotoxic function inside a self pollen tube (Kubo ). The involvement of multiple polymorphic SLF genes in pollen specificity can explain why the first SLF gene identified in P. inflata and Antirrhinum shows a lower degree of allelic sequence diversity than the S-RNase gene, which by itself controls pistil specificity. The PiSLF gene has been renamed type-1 SLF and designated as SLF1, and its alleles are designated as S. For example, PiSLF is designated as S. The additional types of SLF genes are named type-2 SLF (designated as SLF2), type-3 (designated as SLF3), etc. Since this report only deals with PiSLF, we will still use the old name of PiSLF to prevent confusion.

In the canonical SCF (SKP1-CULLIN-1-F-box) complex, the F-box protein interacts with SKP1 through its N-terminal F-box domain, and interacts with its substrate(s) through another protein–protein interaction domain at the C-terminus. The putative PiSLF-containing E3 ligase complex does not appear to contain an SKP1-like protein, but instead contains PiSBP1, which is three times the size of PiRBX1 (the RING-finger component of a conventional SCF complex) and could play the roles of both SKP1 and RBX1 (Hua and Kao 2006). This finding, coupled with the finding that the C-terminal domain (CTD) of PiSLF2 (lacking the F-box domain) can interact with PiSBP1 (Hua and Kao 2006), raises a question as to whether the F-box domain of PiSLF is necessary for its function in SI. To address this question, we constructed a truncated PiSLF gene encoding PiSLF2(CTD), which lacks the predicted F-box domain (amino acids 9–49) and the N-terminal eight amino acids, fused the coding sequence for a GFP (green fluorescent protein) to its 3′ end, and used the pollen-specific LAT52 promoter of tomato (Twell ) to express this transgene in S plants of P. inflata. We wished to determine whether the expression of PiSLF2(CTD) would cause breakdown of SI in S transgenic pollen, as is the case with the full-length PiSLF2. If the F-box domain is not necessary for the function of PiSLF in SI, the transgenic plants should exhibit the same SI behaviour as the transgenic plants expressing the full-length PiSLF2. However, if the F-box domain is required for the function of PiSLF, the over-expression of PiSLF2(CTD) could have a dominant-negative effect on the function of endogenous PiSLF2.

Moreover, contrary to the finding with PiSLF, Qiao suggested that AhSLF2 of Antirrhinum might be a component of a canonical SCF complex. It was subsequently found that a novel class of SKP1-like protein in Antirrhinum, named AhSSK1 ( Antirrhinum hispanicum SLF-interacting SKP1-like1), and in P. hybrida, named PhSSK1, interacted with the F-box domain of certain allelic variants of AhSLF and PhSLF (P. hybrida SLF) (Huang ; Zhao ). In this report, we identified an orthologue of AhSSK1 and PhSSK1 in P. inflata, and tested its interactions with three allelic variants of PiSLF that have been shown to be involved in controlling pollen specificity in SI.

Materials and methods

Plant materials

The S, S, S and S genotypes of P. inflata were described by Ai , and the S genotype of P. inflata was described in Wijsman (1983).

Generation of Ti plasmid constructs and plant transformation

The Ti plasmid construct for PiSLF was generated by a procedure similar to that described for the generation of the Ti plasmid construct for PiSLF (Hua ). The C-terminal domain construct contains a 1017-bp coding sequence of PiSLF, lacking the N-terminal 150 bp that encodes the F-box domain. Polymerase chain reaction (PCR) was performed to amplify the cDNA for PiSLF using PiSLF FOR (5′-AACCATGGCTATCAATCGCAAAACAAAC-3′) and PiSLF REV (5′-GCGGCCGCAAATTTTTGTACTTTTGTAC-3′) primers. The Ti plasmid construct was transformed into Agrobacterium tumefaciens (LBA4404) by electroporation, and Agrobacterium-mediated plant transformation was performed according to the method described in Lee .

Genomic DNA isolation and gel blot analysis

Genomic DNA was isolated from 0.5 g of young leaf tissue with Plant DNAzol Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. Each genomic DNA sample (15 μg) was digested with HindIII (32 units) for 16 h at 37 °C, and the DNA fragments were separated by electrophoresis overnight on 0.7 % (w/v) agarose gels and transferred to positively charged nylon membranes (Biodyne B; Pall, Port Washington, NY, USA) overnight. The 888-bp fragment of PiSLF was used as probe (Sijacic ), and labelled with 32P using the Ready-to-Go Labeling kit (GE Healthcare, Piscataway, NJ, USA). The membranes were treated as described by Skirpan , and then exposed to X-ray film at −80 °C for 24 h with an intensifying screen.

Visualization of GFP fluorescence in pollen tubes

Pollen was collected and germinated as described in Meng . The samples were visualized using a Nikon Eclipse 90i epifluorescence microscope (Nikon, Shinjuku, Japan).

Reverse transcription–PCR analysis

Total RNA was isolated from various plant tissues using Trizol reagent (Invitrogen). For each plant, 5 μg of RNA were used to synthesize cDNA in the presence of SMARTScribe reverse transcriptase according to the manufacturer's protocol (Clontech, Mountain View, CA, USA). In all, 250 ng of RNA, 250 ng of the resulting cDNA and 0.1 μg of genomic DNA were separately used for PCR, with one of the following primer pairs: SLF FOR (5′-AACCATGGCTATCAATCGCAAAACAAAC-3′) and GFP REV (5′-AGGTGGTCACGAGGGTG-3′) for PiSLF; Actin FOR (5′-GGCATCACACTTTCTACAATGAGC-3′) and Actin REV (5′-GATATCCACATCACATTTCATGAT-3′) for the actin gene; PiSSK1 FOR (5′-AAGGATCCATATGGCATCAG-3′) and PiSSK1 REV (5′-CGGAGCTCTAATTGACAGTATCA-3′) for PiSSK1. Polymerase chain reaction was performed with 95 °C denaturation for 3 min followed by 30 cycles of 95 °C for 30 s, 56 °C for 30 s and 72 °C for 1 min for amplifying PiSLF and PiSSK1, or followed by 20 cycles of 95 °C for 30 s, 60 °C for 30 s and 72 °C for 45 s for amplifying the actin gene. A final extension at 72 °C for 10 min was performed after the amplification cycles.

Protein gel blot analysis

For protein blot analysis of anther proteins, total protein was extracted from stage 5 anthers as described by Lee , using protein extraction buffer as described by Hua . Protein concentrations were determined using the BioRad protein assay system (Bio-Rad, Hercules, CA, USA). After quantification, protein samples were resolved on 10 % polyacrylamide gels and transferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA, USA). The membranes were then immunoblotted using rabbit anti-GFP antibody (1:1500 dilution; Abcam, San Francisco, CA, USA) and peroxidase-linked sheep anti-rabbit IgG (1:10 000 dilution; GE Healthcare) to detect GFP and GFP fusion proteins. For protein blot analysis of yeast proteins, yeast cells were harvested between OD600 0.4 and 1.0, resuspended in 20 % trichloroacetic acid (TCA), vortexed by glass beads for 12 min, washed with 5 % TCA and resuspended in 0.5 M Tris, pH 7.4. 3× sodium dodecyl sulphate–polyacrylamide gel electrophoresis buffer was added and boiled for 4–5 min to denature the proteins. The protein blot analysis was carried out as described above, except that the primary antibody was mouse anti-HA (1:1000 dilution; Babco, Princeton, NJ, USA) and the secondary antibody was peroxidase-conjugated goat anti-mouse IgG (1:25 000 dilution; Thermo Scientific, Waltham, MA, USA).

Visualization of pollen tube growth in pollinated pistils

Pollinated pistils for visualization were prepared as described by Fields . The samples were visualized under UV light using a Nikon Eclipse 90i epifluorescence microscope (Nikon).

PCR genotyping and analysis of progeny

Genomic DNA was extracted as described in the section ‘Genomic DNA isolation and gel blot analysis’. For each plant, 100 ng of genomic DNA were used for PCR with primer pairs of SLF FOR (5′-AACCATGGCTATCAATCGCAAAACAAAC-3′) and GFP REV (5′-AGGTGGTCACGAGGGTG-3′) for the PiSLF transgene, and with primers specific to the S and the S genes to analyse the S-genotype of each progeny plant (Meng ). Polymerase chain reaction was performed with 95 °C denaturation for 3 min followed by 30 cycles of 95 °C for 30 s, 55 °C for 30 s, 72 °C for 1 min and a final extension at 72 °C for 10 min.

Screening of the bacterial artificial chromosome library by PCR

The S bacterial artificial chromosome (BAC) library of P. inflata was constructed by McCubbin and screened as described in Wang . Polymerase chain reaction was performed using a pair of primers, PhSSK1 FOR (5′- ATGGCATCAGAAAAG-3′) and PhSSK1 REV (5′-TAATTGACAGTATC-3′), specific to the coding region of PhSSK1, and five positive clones were obtained. Bacterial artificial chromosome DNA was isolated from the positive clones and sequenced using primers PhSSK1 FOR, PhSSK1 REV, F408 (5′-GCCTAACCTGACAATCCACTTT-3′) and R603 (5′-AACCAAAGTGACCAAGCAAAA-3′).

DNA sequence analysis

All DNA sequencing was carried out at the Nucleic Acid Facility of The Pennsylvania State University (http://tanager.huck.psu.edu). Nucleotide sequences were assembled and analysed using DNA Strider 1.2.1. Alignments of amino acid sequences were performed using ClustalW (http://www.ebi.ac.uk/clustalw/); gonnet250 protein weight matrix was selected, and the gap opening and extension parameters were 10 and 0.2, respectively. Alignments were shaded using Boxshade version 3.21 (http://www.ch.embnet.org/software/BOX_form.html).

Yeast two-hybrid assay

The yeast two-hybrid assay was carried out as described in the Matchmaker™ gold yeast two-hybrid system user manual (Clontech). The coding sequences of PiSLF, PiSLF, PiSLF, PiSLF and PiSLF were cloned in-frame with the coding sequence of the GAL4 binding domain (BD) in pGBK-T7. The coding sequence of PiSSK1 was cloned in-frame with the coding sequence of the GAL4 activation domain (AD) in pGAD-T7. PiSBP1 in prey vector pGAD-C1 was previously made (Hua and Kao 2006). To test the interaction between two proteins, the corresponding BD and AD constructs were co-transformed into Saccharomyces cerevisiae Y2HGold (Clontech), and the transformants were plated out on synthetic dropout medium without leucine or tryptophan to select for cells in which both BD and AD fusion proteins were co-expressed. Transformants were streaked out on selective plates lacking adenine, histidine, leucine and tryptophan, and selective plates lacking adenine, histidine, leucine and tryptophan, but containing X-α-Gal and aureobasidin, to examine growth and galactosidase activity.

Results

Generation of transgenic S plants carrying LAT52-PiSLF

We previously used the pollen-specific LAT52 promoter of tomato (Twell ) to express the entire coding sequence of PiSLF fused in-frame to the coding sequences of a 13-amino-acid linker and GFP in S transgenic plants (Hua ). Using the LAT52 promoter to express higher than normal levels of PiSLF2 fused to GFP did not affect the viability of transgenic pollen, nor did it affect the in vivo function of PiSLF2. Thus, we used the same strategy to express PiSLF in which the coding sequence for the N-terminal 49 amino acids, including the predicted 41-amino-acid F-box domain, was deleted from PiSLF. The resulting pBI-LAT52-PiSLF construct was introduced into S plants of P. inflata, and 35 T0 transgenic plants were obtained.

We first examined all the transgenic plants for the presence and copy numbers of the transgene by genomic DNA blot analysis, using as probe an 888-bp fragment of the PiSLF coding region (without the F-box motif region, Fig. 1A) (Sijacic ; Hua ). Seven transgenic plants, designated as PiSLF-9, -11, -18, -25, -28, -30 and -33, carried a single copy of the transgene and the remainder carried two or more copies. The transgene insertion patterns of representative plants are shown in Fig. 1B, along with the results of two wild-type S plants serving as controls.

Fig. 1

Generation of (A) Schematic representation of the full-length PiSLF and the C-terminal domain PiSLF transgene constructs. The full-length construct contains a 1167-bp coding sequence of PiSLF, and the C-terminal domain construct contains a 1017-bp coding sequence of PiSLF, lacking the 5′-terminal 150 bp encoding the F-box domain. Each construct also contains the coding sequence of a GFP fused in-frame to the last codon of the coding sequence. NOS, the gene encoding nopaline synthase; pro, promoter; ter, transcription terminator; NPT II, the gene encoding neomycin phosphotransferase II (conferring kanamycin resistance). The region between the right (RB) and the left border (LB) is integrated into transgenic plants. (B) Genomic DNA gel blot analysis of 13 T0 PiSLF/S transgenic plants and two S wild-type plants. An 888-bp fragment of the PiSLF coding region without the F-box motif was used as probe. Each lane contains 15 μg of genomic DNA digested with HindIII (32 units). Asterisks denote the endogenous PiSLF and PiSLF genes. DNA size markers are indicated.

Expression of PiSLF in the pollen of transgenic plants

We examined the expression of PiSLF in transgenic plants by observing the fluorescence of the GFP-fused protein in in vitro-germinated pollen tubes. For transgenic plants that carry one copy of PiSLF, ∼50 % of the pollen tubes should show fluorescence if the GFP-fused protein is indeed expressed and not toxic to pollen/pollen tubes. All seven transgenic plants showed fluorescence in ∼50 % of their germinated pollen tubes (Table 1 and [Additional Figure 1]), consistent with the finding by genomic DNA blot analysis (Fig. 1B) that they each carried a single copy of the transgene.

Table 1

Examination of GFP-tagged transgene expression in pollen tubes.

Transgenic linea	Total number of pollen tubes examinedb	Total number of fluorescent pollen tubes
9	96/111/117	49/59/63
11	107/109/121	53/59/65
18	115/133/146	61/72/77
25	106/121/127	47/50/59
28	105/111/127	52/57/65
30	120/123/141	51/56/70

aTransgenic line carrying one insertion of the transgene.

bPollen grains from each flower were germinated as described in Results. Five images were taken for each sample, and the total pollen tube number was calculated by summing up all the pollen tubes in the five images. Three flowers were examined for each transgenic line.

To confirm that the fluorescence observed in in vitro-germinated pollen tubes was due to PiSLF2(CTD):GFP, we performed reverse transcription (RT)–PCR on three of the single-copy PiSLF transgenic plants to see whether the transgene transcripts could be detected in their stage 5 anthers (defined in Lee ). Total RNA was isolated from PiSLF-9, -11, -28 and a wild-type S plant, and used for reverse transcription (marked with ‘RT+’). As shown in the upper panel of Fig. 2A, a DNA fragment of the expected size, ∼1.2 kb, was detected in the cDNA samples of all these three transgenic plants (lanes labelled ‘9’, ‘11’ and ‘28’, marked with ‘RT+’), using a pair of transgene-specific primers, but not in the cDNA sample of the wild-type S plant (lane labelled ‘−’, marked with ‘RT+’). Plasmid DNA containing PiSLF (lane labelled ‘P’) also produced a DNA fragment of similar size to that of the corresponding RT–PCR product, suggesting that PiSLF was transcribed in stage 5 anthers of these transgenic plants. As a negative control, genomic DNA of the wild-type S plant (lane labelled ‘G’) did not produce any fragment. The same RNA samples used for RT–PCR were also amplified using a primer pair for actin (lower panel), and the results showed that these samples contained approximately equal amounts of cDNA, suggesting that absence of the RT–PCR band of PiSLF transgene in the sample of wild-type S plant was not due to insufficient amounts of cDNA. The plasmid DNA sample did not produce any fragment, while the genomic DNA sample yielded a fragment larger in size than that of the RT–PCR product (due to the presence of an intron). When PCRs were performed on total RNA from each plant (marked with ‘RT−’), no fragment was detected with either pair of primers, indicating that there was no genomic DNA contamination on the cDNA samples.

Fig. 2

Expression analysis of (A) Reverse trancriptase–PCR analysis of RNA transcripts of PiSLF in stage 5 anthers of transgenic plants. Anthers were separately collected from three PiSLF transgenic plants (lanes labelled ‘9’, ‘11’ and ‘28’) and a wild-type S plant (lane labelled ‘−’). Five micrograms of RNA isolated from each plant were used for reverse transcription, and 250 ng of the resulting cDNA were used for PCR. Each panel shows the results of amplification of the cDNA samples, total RNA, plasmid DNA containing PiSLF (lane labelled ‘P’) and 0.1 μg of genomic DNA of the wild-type S plant (lane labelled ‘G’). A primer pair specific to PiSLF was used in the results shown in the upper panel, and a primer pair specific to the actin gene was used in the results shown in the lower panel. ‘RT+’, PCR performed on cDNA synthesized from total RNA in the presence of reverse transcriptase. ‘RT−’, PCR performed on total RNA in the absence of reverse transcriptase. ‘M’ indicates EcoRI and HindIII digested λ DNA used as size markers. (B) Protein gel blot analysis of PiSLF2(CTD):GFP in stage 5 anthers of transgenic plants. The transgenic plants, as well as wild-type S plants and two previously generated transgenic plants used as controls, are indicated at the top of each lane. Approximately 30 μg of total protein isolated from stage 5 anthers of each plant were used for gel electrophoresis. The protein bands were detected by a rabbit anti-GFP antibody and peroxidase-linked sheep anti-rabbit IgG. The single asterisk indicates the protein band corresponding to the PiSLF2:GFP fusion protein. The double asterisks indicate the band corresponding to the free GFP, and the arrow indicates the band corresponding to the PiSLF2(CTD):GFP fusion protein. Molecular mass markers are shown on the left of the figure.

We next performed protein gel blot analysis, using an anti-GFP antibody, to examine whether the GFP fusion protein was produced in stage 5 anthers (Fig. 2B). Total anther proteins were separately isolated from four PiSLF transgenic plants, a wild-type S plant, and two previously generated transgenic plants, LAT52-GFP/S (Dowd ) and PiSLF (Hua ). A protein band of the expected molecular mass, ∼65 kDa, was detected in stage 5 anthers of all four PiSLF transgenic plants (indicated by the arrow in Fig. 2B), but not in the other plants. The PiSLF transgenic plant was used as a positive control for the GFP fusion protein. As predicted, the band corresponding to the fusion protein PiSLF2:GFP was of a higher molecular mass than the band corresponding to the fusion protein PiSLF2(CTD):GFP. The band corresponding to the free GFP, indicated by the LAT52-GFP/S transgenic plant, was also detected in PiSLF transgenic plants, likely resulting from the cleavage of GFP fusion protein during sample preparations as we had previously observed with PiSLF2:GFP (Hua ; Fields ). Taken together, the fusion protein PiSLF2(CTD):GFP was indeed produced in stage 5 anthers of these transgenic plants at a similar or higher level than that of PiSLF2:GFP produced in the previously generated transgenic plant (Fig. 2B).

Transgenic S plants producing PiSLF2(CTD):GFP in pollen remained self-incompatible

The PiSLF transgenic plant used as control in the protein gel blot experiment was self-compatible, and set large fruits with seed numbers comparable to those obtained from compatible pollination between wild-type plants (Hua ; Fields ; this study). However, when we used pollen from the S transgenic plants that produced PiSLF2(CTD):GFP to pollinate themselves or wild-type S plants, none of the pollination resulted in any fruit set. We compared pollen tube growth in pistils of the same wild-type S plant 16 h post-pollination with pollen from the PiSLF transgenic plant, the PiSLF transgenic plants as well as two wild-type plants, S and S (Fig. 3A). Consistent with the fruit set result, most pollen tubes produced by transgenic plant PiSLF-11 stopped in the upper segment of the pistil, similar to pollen tubes from a wild-type S plant, whereas most pollen tubes from the transgenic plant PiSLF grew through the entire pistil, similar to pollen tubes from a wild-type S plant (Fig. 3A). These results suggested that the lack of fruit set for PiSLF transgenic plants was due to a genuine SI response.

Fig. 3

Self-incompatibility phenotypical analysis of (A) Examination of transgenic and wild-type pollen tube growth in pistils 16 h post-pollination. The same wild-type S plant was used as the female parent in all crosses, and the genotype of the pollen parent is shown at the bottom of each image. Pollen tubes were stained with aniline blue and visualized under an epifluorescence microscope. Only the segment near the bottom of the pistil is shown. (B) Polymerase chain reaction analysis of progeny from pollination of a wild-type S plant by pollen of PiSLF-11. Genomic DNA (∼100 ng) extracted from 16 progeny plants and from three wild-type plants (S, S and S) was used as template for PCR, with a specific primer pair corresponding to the coding sequence of PiSLF (upper panel), and primers specific to S (middle panel) or S (bottom panel). The PCR products from genomic DNA of PiSLF-11 (indicated as T0) and the Ti plasmid used for transformation (indicated as P) are shown as positive controls. M indicates EcoRI and HindIII digested λ DNA used as size markers.

To rule out the possibility that the PiSLF transgene might have affected the viability of the pollen/pollen tubes, we used pollen from three single-copy transgenic plants, PiSLF-9, -11 and -28, to separately pollinate S and S wild-type plants. Large fruits with an average seed number ∼140 per fruit were obtained from all the crosses. We then used PCR to determine the S-genotypes and the inheritance of the transgene in the resulting progenies. Pollination results and progeny analysis are summarized in Table 2. Polymerase chain reaction analysis using S- and S-specific primers showed that all plants in both progenies were S genotype. Polymerase chain reaction analysis using primers specific to the PiSLF transgene showed that ∼50 % of the plants in each progeny carried the transgene. These results suggested that S pollen expressing the transgene and S pollen expressing the transgene were accepted by S and S pistils, respectively. Moreover, S pollen producing PiSLF2(CTD):GFP was rejected by the S pistil, because no progeny plant carrying the transgene was S genotype. Figure 3B shows the results of 16 plants in the progeny from pollination of a wild-type S plant by pollen of the transgenic plant PiSLF-11. Nine of them were wild-type S, and seven of them were S with the transgene. We used pollen from these three transgenic plants to pollinate S wild-type plants (Table 2), and based on PCR analyses similar to those described above, each progeny contained S wild-type, S wild-type, S carrying the transgene and S carrying the transgene in a ratio of ∼1:1:1:1. Thus, the PiSLF transgene did not affect the viability or SI behaviour of S or S pollen.

Table 2

Summary of pollination results testing the effect of PiSLF2(CTD):GFP transgene on SI behaviour and viability of transgenic pollen.

Transgenic plant (male parent)	Genotype of wild-type plant (female parent)	Average number of seeds per fruita	Per cent of progeny inheriting transgeneb	Genotype of progeny
PiSLF2(CTD):GFP/S2S3-9	S2S2	135 ± 10	47 % (15)	S2S3
PiSLF2(CTD):GFP/S2S3-11	S2S2	145 ± 15	50 % (20)	S2S3
PiSLF2(CTD):GFP/S2S3-28	S2S2	130 ± 15	43 % (21)	S2S3
PiSLF2(CTD):GFP/S2S3-9	S3S3	145 ± 15	43 % (21)	S2S3
PiSLF2(CTD):GFP/S2S3-11	S3S3	155 ± 10	57 % (21)	S2S3
PiSLF2(CTD):GFP/S2S3-28	S3S3	145 ± 10	41 % (17)	S2S3
PiSLF2(CTD):GFP/S2S3-9	S6S6	175 ± 10	39 % (18)	S2S6, S3S6
PiSLF2(CTD):GFP/S2S3-11	S6S6	185 ± 10	55 % (22)	S2S6, S3S6
PiSLF2(CTD):GFP/S2S3-28	S6S6	175 ± 15	56 % (16)	S2S6, S3S6

aEach number was calculated from five fruits.

bEach number in parentheses is the total number of plants analysed by PCR.

Identification of PiSSK1 from P. inflata

Huang identified AhSSK1 of Antirrhinum by yeast two-hybrid screens using AhSLF-S2 as bait, and Zhao identified PhSSK1 (P. hybrida SSK1) based on its homology to AhSSK1. PhSSK1 and AhSSK1 share 48.3 % amino acid sequence identity, and both are specifically expressed in pollen. The loss of SI function for PiSLF2(CTD) might be caused by its inability to interact with an orthologue of PhSSK1 in P. inflata without the F-box domain. To identify an SSK1-like protein, we screened the previously constructed S BAC library of P. inflata (McCubbin ) by PCR using primers designed based on the coding sequence of PhSSK1. Sequencing analysis of one of the five clones obtained revealed a 537-bp open reading frame, interrupted by a 496-bp intron. The deduced amino acid sequence of the protein, named PiSSK1, shared 98.9 % identity with that of PhSSK1. Figure 4A shows an alignment of the amino acid sequences of PiSSK1, PhSSK1 and AhSSK1. (A more comprehensive alignment, including several typical plant SKP1-like proteins, is shown in [Additional Figure 2].) Only two amino acids, not among the key residues predicted to be involved in binding an F-box domain (Huang ), are different between PiSSK1 and PhSSK1, suggesting that an SSK1 orthologue is present in P. inflata.

Fig. 4

Identification of an (A) Alignment of the deduced amino acid sequences of PiSSK1 of P. inflata, PhSSK1 of P. hybrida, and AhSSK1 of A. hispanicum. The highest consensus residue for each aligned position is indicated at the bottom. Amino acid residue numbers are indicated. The predicted key residues of AhSSK1 involved in binding F-box domain (Huang ) are underlined. Only two amino acid residues, enclosed in red-lined boxes, are different between PhSSK1 and PiSSK1. (B) Tissue-specific expression profile of PiSSK1. RNA extracted from leaves, styles, anthers of stages 1–5 and in vitro-germinated pollen tubes was used for reverse transcription. In all, 250 ng of the resulting cDNA were used for PCR. Each panel shows the results of amplification of the cDNA samples, 0.1 μg of genomic DNA of a wild-type S plant (indicated as G) and water, using a primer pair specific to PiSSK1 (upper panel) and a primer pair specific to the actin gene (lower panel). The single asterisk indicates a non-specific band amplified using the primers for actin. M indicates EcoRI and HindIII digested λ DNA used as size markers.

We used a pair of PiSSK1-specific primers to examine the expression of PiSSK1 by RT–PCR. Similar to PhSSK1 and AhSSK1, PiSSK1 is expressed in pollen and in in vitro-germinated pollen tubes, but not in leaves or styles (Fig. 4B). As shown in the upper panel, a DNA fragment of the expected size, ∼500 bp, was detected in the cDNA samples of anthers from stages 2–5 and pollen tubes, but not in the cDNA samples of leaves or styles. Genomic DNA of a wild-type plant (lane labelled ‘G’) produced a fragment larger in size than that of the RT–PCR product, due to the presence of the intron. The same cDNA samples were also amplified using a primer pair for actin (lower panel of Fig. 4B), and the result showed that these samples contained approximately equal amounts of cDNA. The genomic DNA sample yielded a fragment larger in size than that of the RT–PCR product, indicating that there was no genomic DNA contamination in the RNA samples. Therefore, based on sequence identity, sequence alignment and gene expression pattern, we concluded that PiSSK1 is a bona fide SSK1 gene.

PiSBP1, but not PiSSK1, interacted with three allelic variants of PiSLF based on yeast two-hybrid assay

Using the yeast two-hybrid assay, Huang showed that AhSSK1 interacted with AhSLF-S2 and AhSLF-S5, but not with AhSLF-S1 or AhSLF-S4. Similarly, PhSSK1 interacted with AhSLF-S2, AhSLF-S5 and PhSLF-Sv, but not with AhSLF-S1 or AhSLF-S4 (Zhao ). It is unclear why only a subset of allelic variants of AhSLF and PhSLF is able to interact with AhSSK1 and PhSSK1 if SSK1 is an integral component of the SLF-containing E3 ligase complex. Moreover, of all the AhSLF and PhSLF proteins identified, only AhSLF-S2 has so far been reported to control pollen specificity (Qiao ).

We used the yeast two-hybrid assay to determine whether PiSSK1 interacts with PiSLF1, PiSLF2 and PiSLF3, all of which had been shown to control pollen specificity (Sijacic ; Kubo ; A. Fields, N. Wang and T.-h. Kao, Penn State University, University Park, USA, unpubl. res.). PiSSK1 was cloned into pGADT7 as a prey and the three alleles of PiSLF were individually cloned into pGBKT7 as baits for co-transformation of yeast cells, and the transformants were plated on selective media to examine protein interactions. As shown in Fig. 5, no interaction was detected between PiSSK1 and any of the three allelic variants of PiSLF. We further used the yeast two-hybrid assay to show that PiSSK1 did not interact with the F-box region of PiSLF2, designated PiSLF2(F-box) (amino acid residues 1–49), or with PiSLF2(CTD).

Fig. 5

Yeast two-hybrid assay of interactions between PiSSK1, PiSBP1 and PiSLFs. The bait (BD fusion) and prey (AD fusion) constructs, as indicated, were introduced into yeast reporter strain Y2HGold. Representative transformants from three independent experiments were streaked out on (A) selective plates lacking adenine, histidine, leucine and tryptophan, and (B) selective plates lacking adenine, histidine, leucine and tryptophan, but containing X-α-Gal and aureobasidin, and were examined for growth and α-galactosidase activity. (C) Summary of the yeast two-hybrid results. + indicates positive interactions observed and − indicates no interactions observed. pGADT7-T (indicated as T) and pGBKT7-53 (indicated as 53) were used as controls.

We previously used one SKP1-like protein of P. inflata as bait for yeast two-hybrid screens of a pollen prey library of S genotype, and found that all the positive clones encode seven different F-box proteins, none of which are encoded by S-locus genes (Hua and Kao 2006). When we used PiSLF2 as bait for yeast two-hybrid screens of the same prey library, all the positive clones identified encode PiSBP1. We then confirmed the interaction between PiSLF2 and PiSBP1 by an in vitro protein-binding assay (Hua and Kao 2006). To rule out the possibility that the lack of interaction between PiSSK1 and the three allelic variants of PiSLF in the yeast two-hybrid assay was due to lack of expression of the latter proteins, we used the same PiSLF1, PiSLF2 and PiSLF3 bait constructs as well as the two truncated PiSLF2 constructs, PiSLF2(F-box) and PiSLF2(CTD), to assay for their interactions with PiSBP1. As shown in Fig. 5, we confirmed the interactions of PiSBP1 with PiSLF2, PiSLF2(F-box) and PiSLF2(CTD), as previously reported (Hua and Kao 2006), and also found that PiSBP1 interacted with PiSLF1 and PiSLF3. PiSBP1 did not interact with the protein encoded by the pGADT7-53 control vector, nor did it interact with PiSSK1 (Fig. 5).

To rule out the possibility that the lack of interaction between PiSSK1 and the three allelic variants of PiSLF was due to lack of synthesis of PiSSK1, we performed protein gel blot analysis to examine whether PiSSK1 was produced in yeast cells carrying pGADT7-PiSSK1. As shown in Fig. 6, an ∼43 kDa band, consistent with the predicted molecular mass of PiSSK1 fused to the GAL4 AD domain and HA (haemagglutinin) epitope tag, was detected only in the total protein extract of yeast cells carrying pGADT7-PiSSK1, suggesting that PiSSK1 was produced in yeast but failed to interact with the three allelic variants of PiSLF.

Fig. 6

Protein gel blot analysis of PiSSK1 expression in yeast. Total proteins were extracted from yeast strains Y2HGold (used in two-hybrid assays), Y187, Y2HGold transformed with pGADT7 (encoding GAL4 AD-HA of ∼18 kDa), Y2HGold transformed with pGADT7-PiSSK1 (encoding GAL4 AD-HA-PiSSK1 fusion protein of ∼40 kDa) and DPY11 transformed with TBP-HA (HA-tagged TATA binding protein of ∼30 kDa) transformed into DPY11. The single asterisk indicates GAL4 AD-HA, the double asterisk indicates GAL4 AD-HA-PiSSK1 fusion protein and the triple asterisk indicates TBP-HA used as positive control. Other bands are yeast proteins that cross-reacted with the primary anti-HA antibody. Molecular mass markers are shown on the left.

Discussion

The F-box domain of PiSLF is required for its function in SI

In this work, we transformed S plants with LAT52-PiSLF to examine whether the F-box domain of PiSLF is required for its SI function in vivo. If the F-box domain is not required, PiSLF transgenic plants should exhibit the same SI behaviour as the PiSLF transgenic plants. We first analysed expression of the GFP-fused protein in in vitro-germinated pollen tubes by their fluorescence, and found that for each of the seven T0 transgenic plants that carried a single copy of the transgene, ∼50 % of the pollen tubes were fluorescent. We chose three of these plants for analyses by RT–PCR and protein gel blotting, and showed that PiSLF was expressed in the transgenic plants and that, in stage 5 anthers, the GFP-fused protein was produced at levels comparable to that of PiSLF2:GFP produced in a previously generated PiSLF transgenic plant. Thus, in these transgenic plants, PiSLF2(CTD):GFP was produced to sufficient levels in pollen for it to function, should it retain the normal function of PiSLF2.

The SI behaviour of the three transgenic plants examined by RT–PCR and protein gel blotting was analysed, and all of them remained self-incompatible, unlike the transgenic plants producing the full-length PiSLF2. Progeny analysis from the crosses between these PiSLF transgenic plants (as male parent) and S and S wild-type plants (as female parent) suggested that the PiSLF transgene did not affect the viability or SI behaviour of S or S pollen. If PiSLF2(CTD) had the same function as PiSLF2, S pollen carrying the transgene would be compatible with S pistil due to competitive interaction, and thus plants of S genotype carrying the transgene would have been obtained from the cross of PiSLF transgenic plants with S wild-type plants. Thus, our results suggest that the F-box domain of PiSLF is required for its function in SI.

Expression of PiSLF does not affect the SI behaviour of transgenic pollen

The LAT52 promoter used to drive the expression of GFP-fused PiSLF2(CTD) is a much stronger promoter than the native PiSLF promoter. Since PiSLF2(CTD) lacking the F-box domain could not function as the full-length PiSLF2, we examined whether over-expression of PiSLF2(CTD):GFP in S pollen would have a dominant-negative effect on the SI phenotype of transgenic plants by out-competing the endogenous PiSLF2 in either assembly into the PiSLF-containing E3 complex, or interaction with the non-self S-RNases inside a pollen tube. In either case, PiSLF2 would be unable to function to target non-self S-RNases for ubiquitination and subsequent degradation, and its essential function in SI would be revealed. Since the specific amino acids of PiSLF and S-RNase involved in their interaction had not been determined and since PiSLF2 shares ∼90 % amino acid sequence identity with PiSLF3, it was also possible that the over-expressed PiSLF2(CTD):GFP could out-compete the endogenous PiSLF3 in S pollen. However, we found that ∼50 % of the plants in each progeny, from crosses between PiSLF(CTD):GFP/S transgenic plants and S or S wild-type plants, carried the transgene, suggesting that over-expression of PiSLF2(CTD):GFP does not affect SI behaviour in either S or S transgenic pollen.

This finding could suggest that PiSLF2(CTD) does not have a dominant-negative effect on endogenous PiSLF2 in S transgenic pollen or on endogenous PiSLF3 in S transgenic pollen. However, this is not likely, because the level of PiSLF2(CTD) produced in the S and S transgenic pollen is much higher than those of endogenous PiSLF2 in S pollen and PiSLF3 in S pollen, and because PiSLF2(CTD) also interacts with PiSBP1 (Fig. 5) and S3-RNase. Rather, the lack of a dominant-negative effect is consistent with the recent finding that in Petunia, multiple types of SLF proteins collaboratively function to recognize the entire suite of non-self S-RNases in pollen (Kubo ). For a given S-haplotype, two or more types of SLF proteins may recognize and target the degradation of a particular non-self S-RNase (Kubo ), and thus loss of function in a single type of SLF gene may not affect the SI behaviour of pollen, because products of some other type(s) of SLF gene(s) may also recognize the same non-self S-RNase targeted by the defective SLF. Therefore, the finding of over-expression of PiSLF2(CTD):GFP is likely because at least one other type of SLF produced in S and S pollen can also interact with S3-RNase and/or S6-RNase to mediate their degradation.

PiSBP1, but not PiSSK1, interacts with PiSLF in yeast two-hybrid assay

Qiao et al. (2004, ) and Huang showed that AhSLF of Antirrhinum might be a component of a typical SCF complex. Contrary to the findings with AhSLF, we previously showed that PiSLF might be in a novel E3 ubiquitin ligase complex that contains PiCUL1-G (a CULLIN-1) and PiSBP1 (a RING-HC protein), but does not contain SKP1 or RBX1 (Hua and Kao 2006). This was based on the findings that (i) PiSBP1 interacts with PiSLF, S-RNase, PiCUL1-G and an E2 ubiquitin-conjugating enzyme; and (ii) PiSLF interacts with S-RNases, but does not interact with any SKP1. According to this model, it is PiSBP1, but not SKP1, that brings PiCUL1-G and PiSLF into the complex. Since PiSBP1 interacts with both male and female specificity determinants, it may play a key role in S-RNase-based SI.

Zhao identified a novel class of SKP-1-like proteins, PhSSK1, in P. hybrida, which is expressed specifically in pollen and might act as an adapter in the SCF complex. They showed that PhSSK1 interacted with AhSLF-S2/-S5 and PhSLF-Sv, and that substantial down-regulation of PhSSK1 led to reduced fertility of cross-compatible pollen, suggesting its involvement in SI. However, it cannot be ruled out that the reduced fertility might be due to lethality to pollen caused by the use of the long-hairpin RNA to suppress the expression of PhSSK1. Xing and Zachgo (2007) reported the pollen lethal phenotype in the arabidopsis RNAi lines generated to suppress AGL18, but did not find the same defective pollen phenotype in the T-DNA knockout mutant. They generated RNAi lines to suppress genes that are not expressed in pollen and demonstrated that the long-hairpin RNA itself, rather than silencing of a specific gene, caused the pollen lethal phenotype. In all the RNAi lines generated, they also observed a lower expression level of the target transcript, and this could be explained by the death of some pollen. Moreover, among the several allelic variants of SLF from A. hispanicum and P. hybrida with which PhSSK1 was found to interact, only the function of AhSLF-S in SI has been demonstrated through its causing the breakdown of SI function in transgenic pollen of a self-incompatible P. hybrida line (Qiao ).

In this work, we identified PiSSK1 in P. inflata, an orthologue of PhSSK1, and examined its interaction with three allelic variants of PiSLF, whose SI function has already been established. None of these proteins interacted with PiSSK1 in the yeast two-hybrid assay. As both AhSSK1 and PhSSK1 were found by the yeast two-hybrid assay to selectively interact with certain allelic variants of AhSLF and PhSLF, we cannot rule out the possibility that PiSSK1 may also interact with some yet untested allelic variants of PiSLF. However, this would raise a question as to what protein serves as the adapter for CULLIN-1 and the allelic variants of SLF that do not interact with SSK1. Thus, although recent studies have suggested the role of a conventional SCFSLF complex in S-RNase degradation (Qiao ; Huang ), questions remain as to the precise biochemical nature of this complex.

Conclusions and forward look

We have shown that in P. inflata, the F-box domain of PiSLF is required for its SI function, and that this requirement is probably not due to the interaction between the F-box domain and a conventional SKP1-like protein (Hua and Kao 2006; Huang ) or a novel SKP1-like protein (SSK1) proposed to be involved in the SCFSLF complex (Huang ). Instead, we have further confirmed the interaction of PiSLF with PiSBP1 (Hua and Kao 2006), suggesting that the F-box domain of PiSLF is involved in the assembly of a PiSLF-containing E3 complex probably through binding to PiSBP1. To definitely establish that PiSBP is a component of this complex, one may isolate the complex, for example by co-immunoprecipitation, and identify its individual components. Given that the pollen specificity determinant comprises multiple types of SLF proteins, it will be of interest to examine whether all of these proteins interact with PiSBP1, and if so, what common domain(s) may be involved in the interaction. PiSBP1 is also expressed in all the vegetative tissues (e.g. leaf, petal, root) examined (Hua and Kao 2006), but its physiological function, if any, in these tissues is unknown. Finally, if the sole function of PiSBP1 in pollen is to serve as a component of the E3 complex involved in degradation of S-RNases, one would expect that specific suppression of its expression in pollen would render the types of SLF proteins that require PiSBP1 for SI function unable to detoxify any S-RNase. As a result, the transgenic pollen would be rejected by pistils of any S-genotype.

Additional information

The following additional information is available in the online version of this article –

Figure 1: Bright field (top) and fluorescence (bottom) images of representative pollen tubes produced by transgenic plant PiSLF-11.

Figure 2: Sequence alignment of PiSSK1 of P. inflata, PhSSK1 of P. hybrida, AhSSK1 of A. hispanicum and several typical plant SKP1-like proteins.

Accession numbers

PISSK1 (JF429902).

Sources of funding

This work was supported by National Science Foundation (USA) grant IOS-0843195 to T.-h.K.

Contributions by the authors

X.M., Z.H. and T.-h.K. designed the experiments. Z.H. made the pBI LAT52 SLF(CTD):GFP construct shown in Fig. 1A and performed some of the plant transformation experiments. P.S. carried out the protein gel blot analysis of PiSSK1 shown in Fig. 6. X.M. performed all the other experiments. X.M. and T.-h.K wrote the manuscript, with input from Z.H. and P.S.

Conflicts of interest statement

None declared.