Non-cell-autonomous regulation of cruciferself-incompatibility by Auxin Response Factor ARF3Titima Tantikanjana1 and June B. Nasrallah1

Department of Plant Biology, Cornell University, Ithaca, NY 14853

Contributed by June B. Nasrallah, October 4, 2012 (sent for review September 17, 2012)

In many angiosperms, outcrossing is enforced by genetic self-incompatibility (SI), which allows cells of the pistil to recognize andspecifically inhibit “self” pollen. SI is often associatedwith increasedstigma-anther separation, amorphological trait thatpromotes cross-pollen deposition on the stigma. However, the gene networks re-sponsible for coordinate evolution of these complex outbreedingdevices are not known. In self-incompatible members of the Brassi-caceae (crucifers), the inhibition of “self”-pollen is triggered withinthe stigma epidermal cell by allele-specific interaction between twohighly polymorphic proteins, the stigma-expressed S-locus receptorkinase (SRK) and its pollen coat-localized ligand, the S-locus cysteine-rich (SCR) protein. Using Arabidopsis thaliana plants that express SIas a result of transformation with a functional SRK–SCR gene pair,we identify Auxin Response Factor 3 (ARF3) as a mediator of cross-talk between SI signaling and pistil development. We show thatARF3, a regulator of pistil development that is expressed in the vas-cular tissue of the style, acts non-cell-autonomously to enhance theSI response and simultaneously down-regulate auxin responses instigma epidermal cells, likely by regulating a mobile signal derivedfrom the stylar vasculature. The inverse correlation we observed instigma epidermal cells between the strength of SI and the levels ofauxin inferred from activity of the auxin-responsive reporter DR5::GUS suggests that the dampening of auxin responses in the stigmaepidermis promotes inhibition of “self” pollen in crucifer SI.

pollen–stigma interaction | receptor signaling | auxin signaling | cell–cellcommunication

Flowering plants having both female and male reproductivestructures within the same flower have evolved a variety of

mechanisms that promote cross-pollination, thereby allowing themto avoid inbreeding depression and maintain genetic variationamong individuals. Among the most recognized outcrossing mech-anisms are physiological self-incompatibility (SI) systems, whichenable cells of the female reproductive tract to recognize and rejectgenetically related pollen grains, and morphological adaptations inflower architecture that increase the separation between stigma andanther heights and thus minimize the chance of physical contactbetween pollen and stigma within a flower (1, 2). Although thesephysiological and morphological barriers to selfing appear to beunrelated mechanistically, they are often found to coevolve, sug-gesting that SI and floral developmental programs are based onintersecting genetic networks.Arabidopsis thaliana is a highly self-fertile species that harbors

nonfunctional alleles of the two genes that determine specificity inthe SI response of theBrassicaceae:SRK, which encodes theS-locusreceptor kinase (SRK), a transmembrane protein displayed on thesurface of stigma epidermal cells, and SCR, which encodes the S-locus cysteine-rich protein, which is the pollen coat-localized ligandfor SRK. In previous studies, we showed that A. thaliana can bemade to express SI by transformation with SRK–SCR gene pairsisolated from self-incompatible members of the Brassicaceae, suchas Arabidopsis lyrata (3–5). In naturally self-incompatible crucifers,the SI response is regulated during stigma development, with SIbeing first evident in mature floral buds just before flower openingand persisting throughout flower development. Similarly, SRK–SCRtransformants of some A. thaliana accessions, such as C24, express

a robust and developmentally stable SI response and these plants donot set seed (4). In contrast, SRK–SCR transformants of otheraccessions, such as Col-0, express transient SI, whereby stigmasdisplay an intense SI response only in mature floral buds and just-opened flowers, but subsequently exhibit breakdown of SI in olderflowers, resulting in abundant seed production (3, 4). Using Col-0 plants transformed with the SRKb–SCRb gene pair isolated fromthe A. lyrata Sb haplotype, henceforth referred to as Col(Sb),we previously demonstrated that a mutation in the RNA-DE-PENDENT RNA POLYMERASE 6 (RDR6) gene drasticallyreduces self-seed production by enhancing two distinct floral char-acters that promote outcrossing, SI, and stigma exsertion (i.e.,stigmas positioned above the anthers) resulting fromenhanced pistilelongation in mature flowers (6).RDR6 functions in TAS1-, TAS2-, TAS3-, and TAS4-derived

transacting small inhibitory RNA (ta-siRNA) biogenesis by me-diating double-strand formation of miRNA-directed cleavageproducts, which are subsequently processed into siRNA duplexesby DICER-LIKE 4 (DCL4) (reviewed in 7). TAS3 ta-siRNA bio-genesis is specifically disrupted by loss-of-function mutations inARGONAUTE 7 (AGO7), an integral component of the spe-cialized RNA-Induced Silencing Complex (RISC) that affectsposttranscriptional cleavage of TAS3 precursor genes (8). Asa result, ago7mutants exhibit up-regulation of the TAS3 ta-siRNAtargets Auxin Response Factors 3 and 4 (ARF3 andARF4) (9–11),which belong to a family of genes encoding DNA-binding proteinsthat bind to auxin response elements in the promoters of auxinresponsive genes (12). We had previously shown that ago7 plantsexhibit enhanced SI and stigma exsertion phenotypes similar tothose observed in rdr6 plants (6). These results suggested thatpositive regulators or effectors of SI and pistil development areregulated by ta-siRNA(s) and that ARF3 and/or ARF4 mightfunction in SI.ARF3, also known asETTIN, had been implicated inpistil development based on the severe pistil malformations ob-served in loss-of-function arf3 mutants (13–15) and on the stigmaexsertion phenotypes observed upon over-expression of ARF3transcripts in wild-type plants (9, 11). Consequently, we examinedthe possibility that RDR6 andAGO7might exert their effect on SIthrough their ARF3 target.

Results and DiscussionOverexpression of ARF3 Enhances the SI Response. We overex-pressed ARF3 in a Col(Sb) transgenic strain by introducing a ge-nomic clone containing a TAS3 ta-siRNA-insensitive mutant ofARF3with its native 5′ and 3′ regulatory sequences (9). Expressionof this nontargeted form of ARF3, hereafter referred to asARF3pr::ARF3mut, is expected to result in increased ARF3 tran-script levels relative towild type, due to increased copy number andto the lack of negative regulation of the ARF3mut transcript byTAS3 ta-siRNA (9).

Author contributions: J.B.N. and T.T. designed research; T.T. performed research; J.B.N.and T.T. analyzed data; and J.B.N. and T.T. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence may be addressed. E-mail: jbn2@cornell.edu or tt15@cornell.edu.

19468–19473 | PNAS | November 20, 2012 | vol. 109 | no. 47 www.pnas.org/cgi/doi/10.1073/pnas.1217343109

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

1

Eighty-three independent Col(Sb)(ARF3pr::ARF3mut) trans-formants were generated and grouped into four classes accordingto their floral phenotypes (Fig. 1A): class I plants (21/83 or 25%

of transformants) produced flowers with normal morphology;class II plants (26/83 or 31% of transformants) produced flowershaving a stigma exsertion phenotype similar to that of rdr6(Sb)and ago7(Sb) flowers, and their stigmas had only partially fusedlobes; class III plants (22/83 or 27% of transformants) producedflowers exhibiting stigma exsertion in combinationwith several otherflower developmental defects, including a more severe incompletefusion of stigma lobes than in class II plants, uncoordinated growthof the carpels at their apical end often causing one stigma lobe to bepositioned higher than the other instead of the congenitally fusedcarpels with synchronized growth observed in wild type, reducedpollen production, and narrow sepals and petals that curled out-ward; and class IV plants (14/83 or 17% of transformants) exhibitedsevere defects in pistil development, including incomplete fusion ofapical ends resulting in multiple stigmas and styles, lack of pollen,and more severe sepal and petal phenotypes than those observed inclass III plants.It should be noted that in an earlier study (9), Col-0 wild-type

plants transformed with the nontargeted ARF3pr::ARF3mut trans-gene used here were reported to exhibit only moderate changes inflower morphology that were restricted to a stigma-exsertionphenotype similar to our class II plants. It is possible that themoresevere developmental defects we observed were due to muchhigher ARF3 transcript levels in our transformants than in thosedescribed in the previous study. Alternatively, SRKb, whichenhances pistil elongation in the rdr6 background (6), may haveacted synergistically with ARF3 to cause the more severe flowerdefects observed in our study.Among Col(Sb)(ARF3pr::ARF3mut) transformants, class II

plants exhibited reduced seed set, whereas class III and class IVplants exhibited total lack of seed set. Reciprocal cross-pollinationsto wild-type Col-0 plants indicated that several factors important forsuccessful pollination were perturbed in these plants. In class IVtransformants, the lack of seed resulted both from the absence ofpollen and the inability of the severely deformed pistils to supportthe growth of wild-type pollen tubes. In class III transformants,very few if any of the pollen grains produced by these plantsgerminated at the stigma surface, even when used to pollinate Col-0 wild-type pistils, indicating that these pollen grains were non-functional. Finally, in some class II transformants, which exhibitedonly a stigma exsertion phenotype, wild-type pollen germinated atthe stigma surface, but pollen tube growth was subsequently abor-ted in the transmitting tract, thus preventing successful fertilization.Despite their floral defects and reduced fertility, Col(Sb)

(ARF3pr::ARF3mut) plants produced well-developed stigma epi-dermal cells similar to those of wild-type flowers, with the excep-tion of class IV plants, which produced underdeveloped stigmaticcells. Consequently, class IV plants will not be considered further.Detailed analyses of compatible pollen–stigma interactions in classI, II, and III plants demonstrated that these transgenic plantsretained normal stigma function. Indeed, manual application ofwild-type pollen grains on the stigmas of these plants showed thesame rates of pollen hydration and pollen tube penetrationthrough the cell wall as observed in wild-type stigmas (Fig. 1B).Thus, the ARF3pr::ARF3mut transgene does not disrupt compat-ible pollen–pistil interactions at the stigma surface.To investigate the effect of ARF3 overexpression on SI, pollen

grains expressing SCRb (hereafter SCRb-pollen) were manuallyapplied to stage 14 stigmas (see Materials and Methods for de-scription of flower developmental stages) of control Col(Sb) plantsand those of class I, II, and III Col(Sb)(ARF3pr::ARF3mut) plants,and pollen tube growth was examined 24 h after pollen application.Col(Sb) stigmas display an intense SI response only in stage 12floral buds and early stage 13 flowers, but subsequently exhibitbreakdown of SI in older flowers (after stage 13) (3). As previouslydescribed (3, 6), stage 14 Col(Sb) stigmas allowed the growth ofnumerous SCRb-pollen tubes that penetrated the wall of stigmaepidermal cells (Fig. 1B). Similarly, SCRb-pollen tubes were not

A

B

Fig. 1. Floral phenotypes and pollen–stigma interactions of Col(Sb) plantsexpressing the ARF3pr::ARF3mut transgene. (A) Floral phenotypes observedin independent transgenic plants. The upper panel shows whole flowersrepresenting each of the four Col(Sb)(ARF3pr::ARF3mut) phenotypic classes:a class I flower with normal morphology, a class II flower with enhanced pistilelongation, a class III flower with enhanced pistil elongation and other flowerdefects, and a class IV flower with severe defects. The lower panel showsphenotypes of the pistil apex exhibiting complete fusion of the stigma ina class I flower, incomplete fusion of the stigma and uncoordinated growth(arrow) of the carpels in class III flowers, and severe defects with multiplestigmas and styles and ectopic formation of ovules (arrowhead) in class IVflowers. (B) Pollen–stigma interactions in transgenic plants. The labels on eachpanel indicate the type of pollination performed, with the stigma parentindicated first and the pollen parent indicated second. The top panel showscompatible interactions of wild-type Col-0 pollen grains on a Col(Sb) stigmaand a class III stigma, with no observed differences in the rates of pollengermination and penetration through the stigma epidermal cell wall after 1 hof pollen application. The middle and bottom panels show the behavior ofSCRb-expressing pollen grains on stage 14 stigmas of Col(Sb) and class III Col(Sb)(ARF3pr::ARF3mut) flowers after 24 h of pollen application. Note thatnumerous SCRb-pollen tubes grow on the Col(Sb) flower. In contrast, SCRb-pollen germination is completely inhibited on the stigmas of class III flowersexhibiting a strong SI response (IIIS), whereas inconsistent SI responses wereobserved between the two lobes of the same stigma in some class III flowersexhibiting weak SI response (IIIW). In the latter flowers, when the pistilexhibited uncoordinated growth at its apical end, a more intense pollen in-hibition was typically observed on the longer side of the carpel than on theshorter side as shown in the bottom panels.

Tantikanjana and Nasrallah PNAS | November 20, 2012 | vol. 109 | no. 47 | 19469

PLANTBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

1

inhibited on the stigmas of class I and class II plants. By contrast,under the same pollination conditions, the stage 14 stigmas of classIII transformants inhibited SCRb-pollen. In one-third (7/22) ofthese transformants, the SI responsewas as intense as that observedin stage 14 stigmas of the stably self-incompatible C24(Sb) plants(4): SCRb-pollen grains typically failed to adhere to the stigmaepidermal cells and hydrate, and as a result, they werewashed awayduring fixation and processing for microscopic examination, andonly callose deposits marking the sites of pollen grain contact withthe stigma epidermal cells were observed (Fig. 1B). Interestingly,class III plants, particularly those exhibiting a relatively weak SIresponse, often produced inconsistent SI responses between thetwo lobes of a single stigma, with one lobe showing strong in-hibition of SCRb-pollen and the other lobe supporting SCRb-pollen germination and pollen tube growth through the cell wall(Fig. 1B). This fluctuation in the SI response within the samestigmaparallels the uncoordinated growth of the two incompletely-fused carpels typically observed in ARF3 overexpressing plants(Fig. 1A) and likely results fromdifferentialARF3 expression in thetwo carpels (see below).Quantitative analysis of ARF3 transcript levels in the various

classes of Col(Sb)(ARF3pr::ARF3mut) plants demonstrated thatthe severity of developmental defects and the enhancement in theSI response were positively correlated with ARF3 mRNA levels(Fig. 2A). Moreover, comparison of SRKb transcript levels in thestigmas of class III plants that exhibited enhanced SI and of class Iplants that exhibited transient SI similar to that observed in Col(Sb) plants demonstrated that SRKb levels in these class III plantswere not increased, but rather slightly reduced, relative to class Istigmas (Fig. 2C). Therefore, the enhanced self-pollen inhibitionobserved in Col(Sb) ARF3 overexpressors was not caused by de-velopmental epistasis that indirectly elevated SRKb levels.Interestingly, the threshold levels of ARF3 transcripts that

affected enhancement of SI were significantly higher in Col(Sb)ARF3 overexpressors than in rdr6(Sb) plants (compare Fig. 2 Aand B). This observation explains why the enhanced SI pheno-type was associated with more severe developmental defects inCol(Sb)(ARF3pr::ARF3mut) flowers (Fig. 1A) than in rdr6(Sb)flowers (6). Loss of RDR6 function is known to result in de-re-pression, not only of ARF3 transcripts but also of transcriptsderived from other ARF genes (16, 17). Consistent with this re-sult, we observed up-regulation of ARF4 transcripts in rdr6(Sb)pistils. It is therefore possible that the enhancement of SI in rdr6(Sb) plants can be affected in the context of relatively low ARF3levels because of cooperative activity of ARF3 with additional ta-siRNA ARF targets.

Expression of Truncated ARF3 Disrupts the SI Response. To confirmthe role of ARF3 in the regulation of SI, we assessed the effect ofloss-of-function arf3 mutants on SI. Different arf3 alleles hadbeen reported to display various degrees of gynoecium de-velopmental defects (13). In particular, a detailed analysis of

T-DNA insertional alleles had shown that arf3 alleles in which theT-DNA insertion was located toward the C terminus, resulting inthe production of a truncated ARF3 protein, exhibited the mostseverely malformed gynoecia, possibly because they dimerizenonproductively with ARF3-interacting proteins (18). Therefore,a Col(Sb) plant was crossed with a plant homozygous for arf3–2,an allele in which the T-DNA insertion was located within exon 10at a site upstream of the TAS3 ta-siRNA target sites (19). Asexpected, F2 plants homozygous for the arf3–2 mutation and car-rying the SRKb–SCRb transgenes produced flowers with mal-formed gynoecia. However, the degree to which the arf3–2 alleleaffected gynoecium development varied significantly even amongdifferent gynoecia produced by the same plant, as previously de-scribed for other arf3 alleles (13). Gynoecia with relatively mildphenotypic defects typically exhibited reduced ovary valves (Fig.3A), whereas gynoecia with the most severe phenotypic defectsmanifested complete loss of the valves and appeared as stalk-likestructures, which nevertheless were capped by stigmas having well-developed epidermal cells (Fig. 3A).Pollination assays of arf3–2(Sb) plants revealed that in mature

stage 12 buds, the stigmas of gynoecia exhibiting reduced valvesretained the intense SI response typically observed in stage 12stigmas of wild-type Col(Sb) (Fig. 3A). In contrast, the stigmas ofvalveless gynoecia exhibited breakdown of SI in these mature buds(Fig. 3A).Notably, this disruptionof SIwas not caused by an indirecteffect of gynoecium malformation on SRKb expression becausea quantitative RT-PCR analysis revealed that SRKb transcriptslevels were not reduced in arf3–2(Sb) stigmas that exhibitedbreakdown of SI relative to those exhibiting SI (Fig. 3B).

ARF3 Acts Non-Cell-Autonomously to Regulate the SI Response. Pre-vious in situ hybridization studies (14, 18) had determined thatARF3 transcripts are not expressed in stigma epidermal cells at anystage offlower development. In pistils at stage 9 through stage 12 offlower development, ARF3 transcripts were detected in the vas-cular strands that lie between phloem and xylem elements. To ex-plain how ARF3 regulates SI despite not being expressed in stigmaepidermal cells, we examined the localization of the ARF3 proteinin pistils in Col(Sb) plants transformed with ARF3pr::ARF3mut-GUS, a previously described construct in which theARF3 promoterdrives expression of a nontargeted ARF3mut-GUS translationalfusion (9). A total of 53 independent Col(Sb) plants expressing theARF3mut-GUS fusion were generated. These plants exhibitedfloral and SI phenotypes similar to theCol(Sb)(ARF3pr::ARF3mut)transgenic plants described above. Analysis of ARF3mut-GUSprotein accumulation in these plants revealed that the levels ofARF3mut-GUS protein, as inferred from the intensity of GUSstaining, were correlated with the severity of the floral phenotype.For example, GUS staining was more intense in class III than inclass I (Sb)(ARF3pr::ARF3mut-GUS) or in Col(Sb) plants (Fig.4A). Consistent with the spatial localization of ARF3 transcripts(14, 18), the ARF3mut-GUS protein was not detected in stigma

A B C Fig. 2. Relative levels ofARF3 and SRKbmRNA in Col(Sb) plants expressing the ARF3pr::ARF3mut trans-gene. (A–B) Relative ARF3 mRNA levels in stage 12pistils of class I, II, and III Col(Sb)(ARF3pr::ARF3mut)plants (A) and of Col-0 and rdr6 plants (B). In A, thephenotypically normal class I plants are used as con-trols for baseline ARF3 transcript levels because theycontain the Sb and ARF3pr::ARF3mut transgenes andare therefore more appropriate controls than wild-type Col-0 in this experiment. (C) Relative SRKbmRNAlevels in stage 12 stigmas of class I Col(Sb)(ARF3pr::ARF3mut) plants, which exhibit breakdown of SI inolder flowers similar to Col(Sb) plants, and of class IIICol(Sb)(ARF3pr::ARF3mut) plants, which exhibit anenhanced SI response in older flowers.

19470 | www.pnas.org/cgi/doi/10.1073/pnas.1217343109 Tantikanjana and Nasrallah

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

1

epidermal cells, but was rather found to accumulate in the stylarvascular tissues below these cells (Fig. 4A). This result was con-firmed by quantitative RT-PCR of stage 12 pistils, which detectedARF3mut-GUS transcripts at appreciable levels in styles but onlyat very low levels in stigmas (Fig. 5A). In class I Col(Sb)(ARF3pr::ARF3mut-GUS) plants, which exhibited a transient SI responsethat broke down in stigmas from stage 14 onward, the ARF3mut-GUS fusion was detected only in young floral buds before stage 12and was not detectable by GUS histological staining in stage 14flowers (Fig. 4A). By contrast, in plants that exhibited strong SI, thisvascular staining of the ARF3mut-GUS fusion persisted after stage13 into the open-flower stage (Fig. 4A). Interestingly, in gynoecia inwhich uncoordinated growth resulted in one of the unfused carpelsbeing longer than the other, ARF3mut-GUS levels typically accu-mulated to higher levels in the longer carpel (Fig. 4A). This differ-ential expression of ARF3 protein likely explains the inconsistentpollination responses often observed in these gynoecia (Fig. 1B). Inany case, the result strongly supports the involvement of ARF3 inthe regulation of pistil elongation and further suggests that eachstylar vascular bundle independently regulates the development andfunction of nearby cells within each half of the pistil.

To define further the cells in which ARF3 acts to regulate the SIresponse, we overexpressed ARF3 specifically in stigma epidermalcells by transforming Col(Sb) plants with an AtS1pr::ARF3mutconstruct, in which expression of the nontargeted ARF3 variant isdriven by the AtS1 promoter, which is active specifically in stigmaepidermal cells (20). Interestingly, none of the 17 Col(Sb)(AtS1pr::ARF3mut) transformants analyzed, even those in which stigmasexpressed 100- to 500-fold higher levels of ARF3 than wild type(Fig. 4B), exhibited an enhanced SI response (Fig. 4B). Therefore,ARF3 transcripts do not affect the SI response in stigma epidermalcells when expressed directly in these cells. Taken together, ourresults indicate that ARF3 acts non-cell-autonomously to regulatethe SI response in stigma epidermal cells through a secondarymobile signal likely derived from the vascular tissue of the style.

ARF3 Acts Non-Cell-Autonomously to Suppress the Auxin-ResponsiveDR5::GUS in Stigma Epidermal Cells. ARF3 has been classified asa transcriptional repressor based on a protoplast transfectionassay that demonstrated its ability to suppress expression ofDR5::GUS, an auxin-responsive reporter gene typically used as

A

B

Fig. 3. Phenotypic and expression analysis of arf3–2(Sb) plants. (A) Theupper panels show pistil phenotypes, which ranged from weak phenotypescharacterized by reduced valves [arf3-2(Sb) weak] to strong phenotypesexhibiting complete loss of the valves [arf3-2(Sb) valveless]. The lower panelsshow interaction of stage 12 stigmas with SCRb-pollen 6 h after pollen ap-plication. A Col(Sb) stigma and an arf3–2(Sb) stigma of a pistil with reducedvalves show complete inhibition of SCRb-pollen, whereas stigmas of valve-less pistils show breakdown of SI and allow the growth of numerous pollentubes. (B) Relative SRKb mRNA levels in the self-incompatible (SI) stigmas ofarf3–2(Sb) flowers having a weak pistil phenotype and exhibiting intense SIin stage 12 stigmas and in the self-compatible (SC) stigmas of arf3–2(Sb)flowers having a strong pistil phenotype and exhibiting breakdown of SI instage 12 stigmas.

A B

C

Fig. 4. Non-cell-autonomous effects of ARF3 on the SI response and on theactivity of the auxin responsive DR5::GUS reporter. (A) Localization of ARF3protein in plants expressing the ARF3pr::ARF3mut-GUS fusion. The left col-umn shows ARF3mut-GUS levels in an inflorescence (top panel) and in thepistils of a young stage 11 bud (11) and a stage 14 flower (14) of a class I Col(Sb)(ARF3pr::ARF3mut-GUS) plant. The right column shows the elevatedARF3mut-GUS protein levels in the inflorescence and the stylar vasculatureof stage 14 flowers of class III Col(Sb)(ARF3pr::ARF3mut-GUS) plants. Notethat in class III flowers exhibiting uncoordinated growth of the carpels, GUSstaining is more intense on the longer side (arrow). (B) Direct expression ofARF3mut in stigma epidermal cells. The top two panels show the results ofRT-PCR of ARF3 transcripts and control actin transcripts in the stigmas of Col(Sb) and three Col(Sb)(AtS1pr::ARF3mut) plants. Numbers between the twogel panels indicate the normalized levels of ARF3 transcripts relative to Col(Sb). The bottom panel shows the profuse growth of SCRb-pollen tubes ona stage 14 stigma of a Col(Sb)(AtS1pr::ARF3mut) plant, demonstrating thatdirect expression of ARF3 in the stigma fails to enhance the SI response inolder flowers. (C) Histochemical GUS staining of stage 14 stigmas fromDR5pr::GUS-expressing class I and class III Col(Sb)(ARF3pr::ARF3mut) plantsand Col(Sb)(AtS1pr::ARF3mut) plants.

Tantikanjana and Nasrallah PNAS | November 20, 2012 | vol. 109 | no. 47 | 19471

PLANTBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

1

a proxy for auxin accumulation (21). Because ARF3 acts non-cell-autonomously to regulate SI in stigma epidermal cells, wereasoned that it might also act non-cell-autonomously to sup-press expression of the DR5:GUS reporter in these cells. Ac-cordingly, we introduced the ARF3pr::ARF3mut construct intoCol(Sb) plants harboring DR5::GUS. As expected, the 50 in-dependent Col(Sb)(DR5::GUS)(ARF3pr::ARF3mut) transformantsgenerated in this manner exhibited a range ofARF3 transcript levelsand fell into the four phenotypic classes described above for Col(Sb)(ARF3pr::ARF3mut) plants. Twenty independent class I trans-formants exhibiting normal flower morphology and 11 independenttransformants exhibiting the floral defects characteristic of class IIIplants were chosen for comparative analysis of DR5-driven GUSexpression in stigmatic cells. In the flowers of Col(Sb) and class Itransformants, GUS activity was evident in stigma epidermal cells(Fig. 4C), and appreciable levels ofGUS transcripts were detected instigma tissue by quantitative RT-PCR (Fig. 5 B and C). By contrast,class III flowers, which accumulate relatively high levels of ARF3(Fig. 2A), exhibited very low, if any, GUS staining in stigma epider-mal cells (Fig. 4C), and correspondingly low levels ofGUS transcriptswere detected in their stigmas (Fig. 5B). However, this suppressiveeffect of ARF3 overexpression on the DR5::GUS reporter was notobserved in AtS1pr::ARF3 transgenic plants that overexpress ARF3directly in stigma epidermal cells (Figs. 4C and 5C).

ConclusionsOur study identifies ARF3 as a modulator of SI and a major TAS3ta-siRNA target responsible for the enhanced SI phenotype wepreviously observed in rdr6(Sb) and ago7(Sb) plants. We founda strict correlation between the level of ARF3 expression and thestrength of the SI response. SI was enhanced when ARF3 tran-scripts were increased due to expression of the ta-siRNA-in-sensitive ARF3mut variant under control of the ARF3 promoter.Conversely, SI was abolished in ARF3 loss-of-function mutants.Importantly, our results suggest that in the pistil apex, ARF3 actsnon-cell-autonomously on stigma epidermal cells by controllingthe production of at least one mobile signal that negativelyregulates SI in these cells. This non-cell-autonomous activity issimilar to that of ARF5/MONOPTEROS (ARF5/MP), a tran-scriptional activator that is expressed in embryonic cells, yetregulates specification of the adjacent extraembryonic hypophy-sis precursor (22–24). In embryonic cells, ARF5/MP activatesauxin transport and the production of a mobile transcriptionfactor, TARGET OF MP5 (TMO7), and it is the accumulationof auxin and TMO7 in the hypophysis precursor that initiates themolecular events leading to formation of the hypophysis.Because ARF3, unlike ARF5/MP, functions as a transcriptional

repressor (21), it might act directly to repress transcription of thegene for the postulated mobile factor. However, it is also possiblethat ARF3 acts indirectly by repressing an activator or a repressor

of this gene. Interestingly, the non-cell-autonomous ARF3-medi-ated down-regulation ofDR5::GUS activation in stigma epidermalcells suggests that ARF3 modulates a second auxin response cas-cade that acts cell-autonomously in these cells. Such interplay be-tween non-cell-autonomous and cell-autonomous auxin responsemodules has been described in hypophysis specification (25). Atpresent, it is not known if the down-regulation of ARF3 down-stream targets and the dampening of a cell-autonomous auxinresponse cascade in stigma epidermal cells are two separate con-sequences of ARF3 activity or if suppression of the proposed mo-bile signal in turn down-regulates the cell-autonomous auxinresponse in the stigma epidermis.The observation that increased ARF3 transcript levels caused

both enhanced SI and pistil aberrations demonstrates that ARF3mediates cross-talk between the pistil development and SI signal-ing pathways. Previous studies had implicated ARF3 in the regu-lation of auxin signaling in the pistil based on the similarity of pistilphenotypes in arf3 mutants and auxin regulatory mutants, such aspinoid, and the fact that treatment of the inflorescence apex withthe polar auxin transport inhibitor naphthylphthalamic acid phe-nocopies arf3mutants (15).Our results support this conclusion andfurther suggest a role for auxin responses in pollen–stigma inter-actions and in the integration of these interactions into the overallpistil development program. Interestingly, in plant–pathogeninteractions, which like SI are based on the ability to discriminatebetween self and nonself, auxin has been implicated in diseasesusceptibility and suppression of auxin signaling in disease re-sistance (26). Future work will determine if a similar situationholds for pollen–stigma interactions, whereby auxin enhancespollen-tube growth at the stigma surface, whereas ARF3-mediatedrepression of auxin signaling in the stigma epidermis promotes or isrequired for “self” pollen rejection.

Experimental ProceduresPlant Materials, Growth Conditions, and Transgenes. Plants of the Col-0 acces-sion carrying A. lyrata SRKb and SCRb, referred to as Col(Sb), were previouslydescribed (2). Transformants were selected onMSmedium (Sigma) containing50 μg/mL hygromycin and 100 μg/mL carbenicillin. All transformation con-structs were introduced into Agrobacterium strain GV3101 and were sub-sequently transferred into Arabidopsis plants by the floral dip method (27).Plants were grown at 22 °C under 16 h light/8 h dark regime.

The nontargeted ARF3pr::ARF3mut and ARF3pr::ARF3mut-GUS constructswere described previously (9) and were transformed directly into Col(Sb)plants. Mutant arf3–2(Sb) plants were selected from F2 plants derived froma cross between an arf3–2 plant and a Col(Sb) plant. To generate Col(Sb)plants containing the DR5::GUS transgene and overexpressing ARF3muttranscripts, a Col(Sb) plant was crossed with a Col-0 plant containing DR5::GUS(28), and F2 plants homozygous for both Sb and DR5::GUS were selected forsubsequent transformation with either ARF3pr::ARF3mut or AtS1pr::ARF3mut.

The AtS1pr::ARF3mut chimeric gene was constructed as follows: A Kpn1genomic fragment containing the coding region of nontargeted ARF3mutand its 3′ regulatory sequence was amplified from the ARF3pr::ARF3mut

A B C

Fig. 5. Quantitative analysis of ARF3 and DR5-driven GUS transcripts in transgenic plants. (A) Rela-tive ARF3 mRNA levels in style (sy) and stigma (st) ofstage 12flowers of class I and class III Col(Sb)(ARF3pr::ARF3mut-GUS) plants. (B and C) Relative GUSmRNAlevels in the stigmas of stage 12 flowers from Col(Sb)(DR5::GUS) plants expressing the ARF3pr::ARF3mut(B) and AtS1::ARF3mut (C) transgenes.

19472 | www.pnas.org/cgi/doi/10.1073/pnas.1217343109 Tantikanjana and Nasrallah

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

1

plasmid (9) using 5′-CTTACGGTACCAAAAGTCATCAAGAAACTCCTCTGAG-3′and 5′-CTTACGGTACCCTTGGATTACGTTTTCTTTGAGTGGCACAGA-3′ primers.The amplified product was digested with Kpn1 and cloned into the Kpn1 siteof a pCAMBIA1300 derivative containing the stigma-specific AtS1 promoterfragment (20). Several clones of the resulting construct were sequenced atthe Cornell University Life Sciences Core Laboratories Center, and one clonethat contained no PCR-generated errors was chosen for introduction intoAgrobacterium and plant transformation.

Pollination Assays. Staging of flower development was based on Smyth et al.(29): Stage 11: Stigma epidermal cells appear at the apex of the gynoeciumand petals reach the top of the short stamens. Stage 12: The final bud stagein which the upper part of the gynoecium becomes fully differentiated, theanthers almost reach their mature length, and the petals reach the top ofthe long stamens. Stage 13: Anthers mature and release pollen grains (an-thesis), petals become visible, and buds open. Stage 14: Long anthers extendabove the stigma. Pollen germination and tube growth were monitored onpollinated stigmas that were treated with fixative (three parts ethanol andone part acetic acid) for at least 30 min, softened in 1 N NaOH at 65 °C for 15min, washed briefly two times in water, and stained in decolorized anilineblue before visualization of pollen tube growth by epifluorescence micros-copy (30). Loss of SI was assessed by pollinating stage 12 flower buds beforeanthesis (29) with pollen grains expressing SCRb, followed by a 2- to 6-hincubation period before processing for microscopy. Enhancement in the SIresponse was assessed by emasculating flower buds before anthesis andallowing them to sit overnight until they opened. The first fully openedflowers (stage 14 flowers) were then pollinated with either wild-type Col-0 pollen grains or Col-0 spollen grains expressing SCRb, after which theywere left for 24 h before fixation.

RNA Analysis. Approximately 30–40 stigmas or styles were dissected frommature stage 12 buds, and total RNA was isolated using the TRIzol reagent(Invitrogen).We treated 1 μgof total RNAwithDNase I (Invitrogen) and reverse-transcribed with oligo(dT) primer and First-Strand cDNA Synthesis Kit for Real-

Time PCR (USB). The resultingfirst-strand cDNAs were subjected to quantitativereal-time PCR using HotStart-IT SYBR Green qPCR Master Mix (2×) (USB) in anApplied Biosystems ViiA 7 Real-Time PCR System. The following gene-specificprimers were used: ARF3: 5′-CAACACTTGTTCGGATGGTG-3′ and 5′-CCCA-CACCAAATGTTCCTCT-3′ (10); GUS: 5′-TCCTACCGTACCTCGCATTACC-3′ and 5′-GACAGCAGCAGTTTCATCAATCAC-3′; SRKb: 5′-AATAACCTGCTCGGCTACGC-3′and 5′-GCTGAATCTACGATGAATGGATCT-3′; ACTIN2: 5′GCACCCTGTTCTTCT-TACCG3′ and 5′AACCCTCGTAGATTGGCACA 3′; and UBIQUITIN CONJUGATINGgene At5g25760: 5′-CTGCGACTCAGGGAATCTTCTAA-3′ and 5′-TTGTGCCATT-GAATTGAACCC-3′. PCR amplification was performed under the following con-ditions: 95 °C for 2min, followed by 40 cycles of 95 °C for 15 s, 55 °C for 30 s, and72 °C for 30 s. Relative transcript levels were determined by the comparative Cт(ΔΔCт) method using the ViiA 7 software, with three replicates of each sample.ARF3 and GUS transcript levels were normalized to the endogenous ACTIN2gene, whereas SRKb transcript levels were normalized to the endogenousUBC gene.

GUS Histochemical Assays. GUS histochemical staining was carried out asdescribed previously (31). Briefly, inflorescences were fixed in 90% (vol/vol)acetone for 20 min at room temperature and washed three times in stainingbuffer (50 mM sodium phosphate buffer pH7, 0.2% Triton X-100) beforetransfer to GUS solution (1.5 mM X-Gluc, 1 mM potassium ferrocyanide,1mM potassium ferricyanide). After vacuum infiltration for 10 min, sampleswere incubated first in GUS solution for 24 h at 37 °C and subsequently ina few changes of 70% (vol/vol) ethanol to clear the tissue of chlorophyll andother plant pigments, and the blue product of the GUS reaction was ex-amined under a dissecting or a light microscope.

ACKNOWLEDGMENTS. We thank James Carrington for the ARF3pr::ARF3-mut and ARF3pr::ARF3mut-GUS constructs, Thomas Guilfoyle for Col-0(DR5::GUS) seed, and Tiffany Crispell and Jensen Lo for technical assistance.The arf3-2 line was obtained from the Arabidopsis Biological Resource Cen-ter (Ohio State University). This article is based upon work supported byNational Science Foundation Grants IOS-0744579 and IOS-1146725.

1. Barrett SC, Jesson LK, Baker AM (2000) The evolution and function of stylar poly-morphisms in flowering plants. Ann Bot (Lond) 85:253–265.

2. Barrett SC (2002) The evolution of plant sexual diversity. Nat Rev Genet 3(4):274–284.3. Nasrallah ME, Liu P, Nasrallah JB (2002) Generation of self-incompatible Arabidopsis

thaliana by transfer of two S locus genes from A. lyrata. Science 297(5579):247–249.4. Nasrallah ME, Liu P, Sherman-Broyles S, Boggs NA, Nasrallah JB (2004) Natural vari-

ation in expression of self-incompatibility in Arabidopsis thaliana: implications for theevolution of selfing. Proc Natl Acad Sci USA 101(45):16070–16074.

5. Boggs NA, Nasrallah JB, Nasrallah ME (2009) Independent S-locus mutations causedself-fertility in Arabidopsis thaliana. PLoS Genet 5(3):e1000426.

6. Tantikanjana T, Rizvi N, Nasrallah ME, Nasrallah JB (2009) A dual role for the S-locusreceptor kinase in self-incompatibility and pistil development revealed by an Arabi-dopsis rdr6 mutation. Plant Cell 21(9):2642–2654.

7. Willmann MR, Endres MW, Cook RT, Gregory BD (2011) The functions of RNA-de-pendent RNA polymerases in Arabidopsis. Arabidopsis Book 9:e0146.

8. Montgomery TA, et al. (2008) Specificity of ARGONAUTE7-miR390 interaction anddual functionality in TAS3 trans-acting siRNA formation. Cell 133(1):128–141.

9. Fahlgren N, et al. (2006) Regulation of AUXIN RESPONSE FACTOR3 by TAS3 ta-siRNAaffects developmental timing and patterning in Arabidopsis. Curr Biol 16(9):939–944.

10. Garcia D, Collier SA, Byrne ME, Martienssen RA (2006) Specification of leaf polarity inArabidopsis via the trans-acting siRNA pathway. Curr Biol 16(9):933–938.

11. Hunter C, et al. (2006) Trans-acting siRNA-mediated repression of ETTIN and ARF4regulates heteroblasty in Arabidopsis. Development 133(15):2973–2981.

12. Guilfoyle TJ, Hagen G (2001) Auxin response factors. J Plant Growth Regul 10:281–291.

13. Sessions RA, Zambryski PC (1995) Arabidopsis gynoecium structure in the wild and inettin mutants. Development 121(5):1519–1532.

14. Sessions A, et al. (1997) ETTIN patterns the Arabidopsis floral meristem and re-productive organs. Development 124(22):4481–4491.

15. Nemhauser JL, Feldman LJ, Zambryski PC (2000) Auxin and ETTIN in Arabidopsis gy-noecium morphogenesis. Development 127(18):3877–3888.

16. Peragine A, Yoshikawa M, Wu G, Albrecht HL, Poethig RS (2004) SGS3 and SGS2/SDE1/RDR6 are required for juvenile development and the production of trans-acting siR-NAs in Arabidopsis. Genes Dev 18(19):2368–2379.

17. Williams L, Carles CC, Osmont KS, Fletcher JC (2005) A database analysis methodidentifies an endogenous trans-acting short-interfering RNA that targets the Arabi-dopsis ARF2, ARF3, and ARF4 genes. Proc Natl Acad Sci USA 102(27):9703–9708.

18. Pekker I, Alvarez JP, Eshed Y (2005) Auxin response factors mediate Arabidopsis or-gan asymmetry via modulation of KANADI activity. Plant Cell 17(11):2899–2910.

19. Okushima Y, et al. (2005) Functional genomic analysis of the AUXIN RESPONSEFACTOR gene family members in Arabidopsis thaliana: Unique and overlappingfunctions of ARF7 and ARF19. Plant Cell 17(2):444–463.

20. Dwyer KG, et al. (1994) A superfamily of S locus-related sequences in Arabidopsis:Diverse structures and expression patterns. Plant Cell 6(12):1829–1843.

21. Tiwari SB, Hagen G, Guilfoyle T (2003) The roles of auxin response factor domains inauxin-responsive transcription. Plant Cell 15(2):533–543.

22. Lau S, De Smet I, Kolb M, Meinhardt H, Jürgens G (2011) Auxin triggers a geneticswitch. Nat Cell Biol 13(5):611–615.

23. Schlereth A, et al. (2010) MONOPTEROS controls embryonic root initiation by regu-lating a mobile transcription factor. Nature 464(7290):913–916.

24. Weijers D, et al. (2006) Auxin triggers transient local signaling for cell specification inArabidopsis embryogenesis. Dev Cell 10(2):265–270.

25. Rademacher EH, et al. (2012) Different auxin response machineries control distinctcell fates in the early plant embryo. Dev Cell 22(1):211–222.

26. Navarro L, et al. (2006) A plant miRNA contributes to antibacterial resistance by re-pressing auxin signaling. Science 312(5772):436–439.

27. Clough SJ, Bent AF (1998) Floral dip: A simplified method for Agrobacterium-medi-ated transformation of Arabidopsis thaliana. Plant J 16(6):735–743.

28. Ulmasov T, Murfett J, Hagen G, Guilfoyle TJ (1997) Aux/IAA proteins repress expres-sion of reporter genes containing natural and highly active synthetic auxin responseelements. Plant Cell 9(11):1963–1971.

29. Smyth DR, Bowman JL, Meyerowitz EM (1990) Early flower development in Arabi-dopsis. Plant Cell 2(8):755–767.

30. Kho YO, Baer J (1968) Observing pollen tubes by means of florescence. Euphytica 17:298–302.

31. Kim K-W, Franceschi VR, Davin LB, Lewis NG (2006) β-Glucuronidase as reporter gene:Advantages and limitations. Arabidopsis Protocols, eds Salinas J, Sanchez-Serrano JJ(Humana Press, Totowa, NJ), pp 263–273.

Tantikanjana and Nasrallah PNAS | November 20, 2012 | vol. 109 | no. 47 | 19473

PLANTBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 5,

202

1