ABCG26-Mediated Polyketide Trafficking and ...ABCG26-Mediated Polyketide Trafficking and Hydroxycinnamoyl Spermidines Contribute to Pollen Wall Exine Formation in ArabidopsisW Teagen

ABCG26-Mediated Polyketide Trafficking andHydroxycinnamoyl Spermidines Contribute to Pollen WallExine Formation in Arabidopsis W

Teagen D. Quilichini, A. Lacey Samuels, and Carl J. Douglas1

Department of Botany, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada

Pollen grains are encased by a multilayered, multifunctional wall. The sporopollenin and pollen coat constituents of the outer pollenwall (exine) are contributed by surrounding sporophytic tapetal cells. Because the biosynthesis and development of the exineoccurs in the innermost cell layers of the anther, direct observations of this process are difficult. The objective of this study was toinvestigate the transport and assembly of exine components from tapetal cells to microspores in the intact anthers of Arabidopsisthaliana. Intrinsically fluorescent components of developing tapetum and microspores were imaged in intact, live anthers usingtwo-photon microscopy. Mutants of ABCG26, which encodes an ATP binding cassette transporter required for exine formation,accumulated large fluorescent vacuoles in tapetal cells, with corresponding loss of fluorescence on microspores. These vacuolarinclusions were not observed in tapetal cells of double mutants of abcg26 and genes encoding the proposed sporopolleninpolyketide biosynthetic metabolon (ACYL COENZYME A SYNTHETASE5, POLYKETIDE SYNTHASE A [PKSA], PKSB, andTETRAKETIDE a-PYRONE REDUCTASE1), providing a genetic link between transport by ABCG26 and polyketide biosynthesis.Genetic analysis also showed that hydroxycinnamoyl spermidines, known components of the pollen coat, were exported fromtapeta prior to programmed cell death in the absence of polyketides, raising the possibility that they are incorporated into the exineprior to pollen coat deposition. We propose a model where ABCG26-exported polyketides traffic from tapetal cells to form thesporopollenin backbone, in coordination with the trafficking of additional constituents, prior to tapetum programmed cell death.

INTRODUCTION

Mature pollen grains are surrounded by a specialized and complexcell wall that serves critical functions in grain survival, stigmaticrecognition, and hydration. The outer pollen wall (exine) is com-posed of a sculptured and structurally rigid sporopollenin back-bone, which in dry stigma species is additionally covered bya lipid-based pollen coat or tryphine (Piffanelli et al., 1998; Hsiehand Huang, 2007). The components of the outer pollen wall aresynthesized by secretory tapetal cells, which form the innermostcell layer of the sporophytic anther wall. The sporopollenin com-ponent of the exine assembles over a short window of develop-mental time, beginning after microspore meiosis and the release offree microspores from callose-bound tetrads, nearing completionby the first pollen mitotic division, and concluding prior to tapetumprogrammed cell death and the second mitotic division (Piffanelliet al., 1998; Hsieh and Huang, 2007; Quilichini et al., 2014a). Assporopollenin deposition progresses, tapetal cells accumulate anarray of lipids, proteins, flavonoids, and phenolic spermidineconjugates, primarily in storage organelles called tapetosomesand elaioplasts, destined for the pollen coat in Arabidopsis thali-ana (Piffanelli et al., 1998; Hsieh and Huang, 2007; Grienenbergeret al., 2009; Quilichini et al., 2014a). After sporopollenin deposition

into the backbone of the exine is complete, tapetum programmedcell death releases lipid-rich contents into the locule, filling theexine crevices to form the pollen coat (Quilichini et al., 2014a, 2014b).Elucidating the chemical composition of sporopollenin has posed

great challenges due to its extreme recalcitrance to degradation.Based on biochemical analyses, sporopollenin is thought to containa mixture of phenolics and aliphatic derivatives (Prahl et al., 1986;Guilford et al., 1988; Rozema et al., 2001; Ahlers et al., 2003;Descolas-Gros and Schölzel, 2007). Genetic approaches, primarilyin Arabidopsis and rice (Oryza sativa), have revealed a number ofgenes encoding enzymes and proteins required for sporopolleninbiosynthesis and deposition, providing clues regarding the com-position and mechanism of sporopollenin biosynthesis and sub-stantiating the key role of the tapetum as the source of sporopolleninprecursors (reviewed by Ariizumi and Toriyama [2011] and Quilichiniet al. [2014b]). Among the Arabidopsis genes required for sporo-pollenin formation, a number encode enzymes with characterizedbiochemical activities, including MALE STERILITY2 (Aarts et al.,1997; Chen et al., 2011), ACYL COENZYME A SYNTHETASE5(ACOS5) (de Azevedo Souza et al., 2009), two POLYKETIDESYNTHASES, PKSA/LAP6 and PKSB/LAP5 (Dobritsa et al., 2010;Kim et al., 2010), TETRAKETIDE a-PYRONE REDUCTASE1 (TKPR1/DRL1) and TKPR2/CCRL6 (Tang et al., 2009; Grienenberger et al.,2010), and two CYTOCHROME P450s (CYP703A2/DEX2 andCYP704B1) (Morant et al., 2007; Dobritsa et al., 2009). Based ongenetic and in vitro biochemical analyses, ACOS5, PKSA, PKSB,and TKPR1 are proposed to function together in the synthesis ofhydroxylated tetraketide a-pyrones, polyketides that may formthe major constituent of sporopollenin (Grienenberger et al., 2010).In support of this model, these enzymes are localized at the tapetum

1Address correspondence to [email protected] author responsible for distribution of materials integral to the findingspresented in this article in accordance with the policy described in theInstructions for Authors (www.plantcell.org) is: Carl J. Douglas ([email protected]).W Online version contains Web-only data.www.plantcell.org/cgi/doi/10.1105/tpc.114.130484

The Plant Cell, Vol. 26: 4483–4498, November 2014, www.plantcell.org ã 2014 American Society of Plant Biologists. All rights reserved.

endoplasmic reticulum, interact in vivo, and have been termed thesporopollenin metabolon (Lallemand et al., 2013).

The presence of phenolic constituents in sporopollenin is welldocumented (Wehling et al., 1989; Gubatz et al., 1993; Ahlers et al.,1999; Weng et al., 2010), but there is little information on the exactenzymes required for these constituents or the specific nature ofthese constituents. Manipulation of the Arabidopsis monolignolbiosynthetic pathway, resulting in enhanced levels of 5-hydrox-yconiferyl alcohol subunits in lignin, leads to abnormalities in pollenwall formation, revealing the joint contribution of certain phenyl-propanoids to lignin and sporopollenin biosynthesis (Weng et al.,2010). Mutations in the TRANSPARENT TESTA4 (TT4) and4-COUMARATE:COENZYME A LIGASE (4CL3) genes, encodingenzymes involved in flavonoid biosynthesis, and in SPERMIDINEHYDROXYCINNAMOYL TRANSFERASE (SHT ), encoding anacyltransferase involved in conjugating spermidine to hydrox-ycinnamic acids, affect the composition of the Arabidopsispollen wall (Grienenberger et al., 2009; Dobritsa et al., 2011).

While the understanding of sporopollenin biosynthesis and com-position has advanced rapidly recently, mechanisms for sporopol-lenin trafficking from the tapetum and exine assembly into thehighly patterned pollen wall remain poorly understood (Ariizumiand Toriyama, 2011; Quilichini et al., 2014b). The spatial separa-tion of tapetal cells from pollen grains in taxa with secretory tapeta(including Arabidopsis and rice) suggests a critical role for theexport of sporopollenin components into the locule during exineformation. ABCG26, an Arabidopsis ATP binding cassette trans-porter, is required for sporopollenin accumulation and has beenproposed to traffic sporopollenin components out of tapetal cellsfollowing tetrad release (Quilichini et al., 2010; Xu et al., 2010; Choiet al., 2011; Dou et al., 2011; Kuromori et al., 2011). The apparentortholog of ABCG26 in rice, ABCG15, is also required for exineformation, suggesting a conserved function for these transportersin sporopollenin export (Niu et al., 2013; Qin et al., 2013). However,the substrates of these ATP binding cassette transporters areunknown, and hypothetical links to sporopollenin traffic remainuntested experimentally.

Once sporopollenin components are exported to the locule,they must be delivered through the aqueous medium to the mi-crospore. Small granules formed along the locule side of tapetalcells, called orbicules or Ubisch bodies, have been proposed tomediate sporopollenin traffic between the tapetum and developingmicrospores in some species (Wang et al., 2003). Orbicules arepresent almost exclusively in species with secretory tapeta but arecuriously absent in Arabidopsis (Huysmans et al., 1998). While theabsence of orbicules in some taxa with secretory tapeta and thepersistence of orbicules after pollen wall formation makes theirproposed function in sporopollenin transport subject to debate,the conspicuous presence of orbicules outside tapetal cells ofsome species suggests that additional mechanisms for the exportof pollen wall constituents may be in place.

In this study, the biosynthesis and export of exine componentswas investigated using a novel approach to visualize living tapetalcells in intact anthers of Arabidopsis. Using a variety of mutants ingenes required for exine biosynthesis and deposition, we identi-fied a genetic link between the polyketide sporopollenin precursorpathway and ABCG26, suggesting that the substrate of ABCG26is a product of the polyketide synthesis metabolon. Additionally,

we identified novel hydroxycinnamoyl (HC) spermidine-containingorbicule-like bodies (ORBs) that accumulate prior to tapetumprogrammed cell death only when the polyketide pathway isblocked. The accumulation of these ORBs in double mutants withabcg26 indicates that additional mechanisms of export for com-ponents of the exine exist and that cotrafficking of polyketides andHC spermidines from tapetal cells may occur. These data indicatethat, rather than all pollen coat components being released byprogrammed cell death, HC spermidines are deposited earlier withcomponents of sporopollenin from metabolically active tapetal cells.

RESULTS

Tapetal Cells of abcg26 Anthers Accumulate IntrinsicallyFluorescent Compounds

To test the hypothesis that the ABCG26 transporter exportssporopollenin precursors from anther tapetal cells, abcg26 mu-tants were examined for the accumulation of exine componentsin the tapetum. Since the tapetum is deep in the anther, imagingthese cells is a challenge. Contradictory results for tapetum ul-trastructure in abcg26 mutants have been reported using trans-mission electron microscopy (TEM) (Quilichini et al., 2010; Choiet al., 2011; Dou et al., 2011; Kuromori et al., 2011), which couldbe due to the extraction of metabolites during fixation (Palade,1952; Morgan and Huber, 1967; Bullock, 2011) and/or the chal-lenges associated with anther sample preparation. While previousreports did not identify accumulations inside abcg26 tapetal cells,some found accumulations behind tapetal cells or reported en-larged tapetal cells (Choi et al., 2011; Dou et al., 2011). To ex-amine intact anthers without fixation, we devised a two-photon(2-P) microscopy method, which enables deep imaging of theinnermost cell layers of the anther in living specimens. In 2-Pmicroscopy, a variation of multiphoton microscopy, simultaneousexcitation by two photons of red-shifted light (700 to 1000 nm) isused to excite fluorophores in the UV range (Zipfel et al., 2003).With 2-P microscopy, all anther cell types from wild-type andmutant plants exhibited intrinsic fluorescence, with emission from420 to 460 nm (pseudocolored teal) and 495 to 540 nm (pseu-docolored red) (Figures 1A to 1H). These wavelength ranges re-flect the emissions from low-wavelength-emitting compoundssuch as hydroxycinnamic acids and from higher wavelength-emitting compounds such as chlorophyll (Cerovic et al., 1999;Buschmann et al., 2000).Given that UV excitation results in strong intrinsic fluorescence

of pollen exine and the numerous reports of phenolic compoundsin sporopollenin, an attractive hypothesis is that ABCG26 trans-ports a compound with autofluorescent properties under UV ex-citation and that this compound would accumulate in the tapetumin the absence of transport. Emission differences were observedbetween wild-type (Columbia-0) and abcg26 (abcg26-1) develop-ing microspore walls (Figures 1A to 1D) and tapetal cells (Figures1E to 1H) and were first visible at the free uninucleate microsporestage (anther stage 8), following tetrad release. Wild-type micro-spores exhibited strong pollen wall fluorescence in both channels(Figure 1A, enlarged in Figure 1B; Supplemental Movie 1),while pollen wall fluorescence was absent from most abcg26

4484 The Plant Cell

microspores (Figure 1C, enlarged in Figure 1D; Supplemental Movie2). Variation in the abcg26 phenotype was common, and some lo-cules contained microspores with a thin layer of pollen wall fluores-cence, which lacked reticulate patterning (Supplemental Figures 1Aand 1C). The loss of intrinsic fluorescence in abcg26microspore wallsis consistent with the absence of an exine wall that has been reportedpreviously in fixed, sectioned material (Quilichini et al., 2010).

In the tapetum, striking differences in intrinsic autofluorescencewere observed, as wild-type tapetal cells emitted low diffusefluorescence with small nonfluorescent vacuoles (Figures 1A and1E, arrows in Figures 1B and 1F; Supplemental Movie 1), whiletapetal cells of abcg26 contained large brightly fluorescent vacuoles

(Figures 1C and 1G, arrows in Figures 1D and 1H; SupplementalMovie 2) with one or more fluorescent cores. Fluorescence emissionof the tapetum vacuole cores was strong in the lower wavelengthrange, and similar structures were not observed in any wild-typetapetum vacuoles. A negative correlation between emission frommicrospores and tapetal cell vacuoles was observed among anth-ers from the same abcg26 plant at comparable stages of de-velopment (Supplemental Figure 1). Anther specimens with highlevels of fluorescence in the tapetum vacuole lacked or had dimfluorescence around microspores (Supplemental Figures 1E and 1F),while specimens with low levels of tapetum vacuole fluorescencehad bright fluorescent emission frommicrospore walls (Supplemental

Figure 1. Vacuolar Inclusions Are Present in abcg26 Mutant Tapetal Cells.

(A) to (H) 2-P microscopy of live wild-type and mutant anthers showing short-wavelength autofluorescence in teal color.(I) to (L) TEM images of cryofixed anthers.(A) and (B) Wild-type anthers showing bright intrinsic fluorescence emitted by microspore walls in the locule and diffuse intrinsic fluorescence in tapetalcells (arrow in the enlargement in [B]).(C) and (D) abcg26 anthers showing locules devoid of fluorescence signal but with bright autofluorescence in tapetal cells (arrows in the enlargement in [D]).(E) and (F)Wild-type anthers focused on the tapetal cell layer showing intrinsic fluorescence emitted diffusely by tapetal cells (arrows in the enlargementin [F]) and the absence of fluorescence emission from tapetum vacuoles. Bright signal was seen from underlying microspore walls (asterisks in [F]).(G) and (H) abcg25 anthers focused on the tapetal cell layer, which contained autofluorescent vacuoles (arrows in the enlargement in [H]) with brightlyemitting punctae not seen in the wild type.(I) and (J)Wild-type anthers showing tapetal cell vacuoles that contained little to no internal contents. (J) shows an enlargement of the boxed area in (I).(K) and (L) abcg26 anthers showing tapetal cell vacuoles containing inclusions (arrows in the enlargement in [L]) within a vacuole filled with electron-dense material.Ep, epidermis; Lo, locule; ML, middle layer; Msp, microspore; Tp, tapetum. Bars in (A), (C), (E), and (G) = 20 mm; bars in (B), (D), (F), (H), (I), and (K) =5 mm; bars in (J) and (L) = 2 mm.

Polyketide Trafficking in Exine Formation 4485

Figures 1A to 1D). In addition to tapetum vacuolar inclusions, brightfluorescent accumulations were often observed on the outerpericlinal face of tapetal cells (arrows in Supplemental Figures1A and 1C).

To complement the 2-P microscopy, anthers were high-pres-sure frozen and freeze substituted (Quilichini et al., 2014a), a pro-tocol that immobilizes cellular metabolites during fixation for TEM.While previous TEM studies failed to identify enlarged vacuoleswith associated inclusions in abcg26 tapetal cells, with the insightprovided by live cell imaging data, careful examination of abcg26tapetal cells by TEM revealed the presence of enlarged vacuolesin cryofixed abcg26 samples. The abcg26 tapetum vacuole coresobserved by TEM were visible as containing one or more circularelectron-dense core structures (Figures 1K and 1L; SupplementalFigures 1G and 1H). Similar core structures were not observed inwild-type tapetal cell vacuoles (Figures 1I and 1J). The cells of themiddle layer of abcg26 anthers at a similar developmental stagealso contained vacuolar accumulations, with the same morpho-logical and staining characteristics as the inclusions in tapetal cells(Supplemental Figures 1I and 1J), and these were never observedin wild-type middle layer cells (Figure 1I).

To test if the distribution of lipidic constituents might differ inabcg26 anthers relative to the wild type, anthers were stained withNile Red, a dye that does not fluoresce in water but becomesfluorescent in hydrophobic environments and organic solvents(Greenspan et al., 1985). Unstained anthers emitted little to nofluorescence in the channel used to capture Nile Red fluores-cence (Supplemental Figures 2A to 2D). Penetration of the Nile Reddye into anther locules required 20 to 30 min, and all images shownwere captured within this time interval (Supplemental Figures2E to 2L). Wild-type microspores exhibited staining in the wall(Supplemental Figure 3A). By contrast, abcg26 microspores,when present, typically exhibited internal staining within 20 min,suggesting altered permeability to the dye, but lacked significant NileRed fluorescence from microspore walls (Supplemental Figure 3C).Wild-type and abcg26 tapetal cells stained with Nile Red revealedabundant lipids in puncta, presumably elaioplasts and tapetosomes,throughout the cells (Supplemental Figures 3B, 3D, 3F, and 3H).Although differences in tapetum lipids were not observed betweenmutants and the wild type, the autofluorescent cores within abcg26tapetum vacuoles (Supplemental Figures 3E, 3G, and 3I) stained withthe hydrophobic dye (Supplemental Figures 3F and 3H) and lackedemission in the Nile Red emission channel when unstained(Supplemental Figure 3J). Wild-type tapetum vacuoles lacked NileRed emission, were smaller than abcg26 vacuoles, and lacked theassociated stained inclusions (Supplemental Figures 3A and 3B).

The development of these 2-P imaging methods for live anthersrevealed that abcg26mutants accumulate enlarged autofluorescentvacuoles with dense core structures. As sporopollenin is lacking onabcg26microspores, these accumulations appear to be composedof sporopollenin components in a pretrafficked state prior to poly-merization on the microspore surface.

The Polyketide Synthesis Pathway Is Required for Inclusionsin the Tapetal Cells of abcg26

The nature of the vacuolar material in the tapetum of abcg26was investigated by examining double mutants affecting both

the ABCG26 transporter and key enzymes required for sporopol-lenin biosynthesis. In a model proposed by Grienenberger et al.(2010) and Quilichini et al. (2014b), ACOS5, CYP703A2, CYP704B1,PKSA and PKSB, and TKPR1 work sequentially to synthesize ahydroxylated tetraketide a-pyrone(s), a major building block ofsporopollenin, since mutants in the genes encoding these enzymesare severely compromised in exine deposition. acos5, pksa pksb,and tkpr1 mutants exhibit highly similar phenotypes to abcg26,including the absence of pollen exine and strong reductions in malefertility. We hypothesized that the product(s) of this pathway, re-ferred to herein as the polyketide pathway, could be exportedby ABCG26. To test this hypothesis, we examined anthers fromdouble or triple mutants of abcg26 and acos5, pksa pksb, and tkpr1using 2-P microscopy and compared these images with theabcg26 single mutant. The fluorescent tapetum vacuoles observedin the abcg26 single mutant (Figures 1H and 1L) were absent in thewild-type control (Figures 2A and 2B) and in the double or triplemutants abcg26 acos5 (Figures 2C and 2D), abcg26 pksa pksb(Figures 2E and 2F), and abcg26 tkpr1 (Figures 2G and 2H). Bycontrast, the tapetum vacuolar inclusions in abcg26 mutants per-sisted in double mutant combinations with other pollen wall mu-tants, including tt4 and 4cl3 (Supplemental Figure 4). These dataindicated that abcg26 vacuolar inclusions were lost only and spe-cifically in polyketide pathway mutants. Because the fluorescentvacuolar material that accumulates in abcg26 requires the tapetum-localized polyketide pathway, these data support the hypothesisthat the substrate of ABCG26 is a polyketide pathway product.

ORBs Accumulate in Polyketide PathwayBiosynthetic Mutants

When examined by 2-P microscopy under UV excitation, thedouble and triple mutants formed by abcg26 in combination withacos5, pksa pksb, and tkpr1 mutants exhibited unexpectedfluorescent punctae surrounding tapetal cells (Figures 2D, 2F, and2H) and some microspores (Figures 2C, 2E, and 2G) that were notobserved in the wild type or abcg26 single mutants (Figures 1Cand 1D). This observation led us to examine the acos5 singlemutant anthers by 2-P microscopy under UV excitation to de-termine if these fluorescent bodies were similarly present in mu-tants of the polyketide pathway. As in the acos5 abcg26 doublemutant, analysis of the acos5 single mutant revealed equivalentpunctate fluorescent bodies that were not observed in wild-typeanthers (Figure 3). The fluorescent puncta were found on all sidesof the tapetum (Figures 3C and 3E), particularly on the outerpericlinal surface of the tapetum (Figures 3D and 3F) and sur-rounding some microspores (Figure 3C). These bodies appearedin acos5 anthers at the uninucleate microspore stage (antherstage 8) and became more abundant around tapeta over thecourse of microspore mitosis (anther stages 9 to 11).These ORBs resembled the orbicules that surround tapetal

cells in many other flowering plant species, including rice, butare absent in Arabidopsis (Wang et al., 2003). AutofluorescentORBs were also observed in anthers of the other polyketidepathway mutants, cyp703a2, pksa pksb, and tkpr1 (SupplementalFigures 5C to 5H), but not in the wild type (Supplemental Figures5A and 5B), and have not been reported previously in descriptionsof these mutants. TEM analysis confirmed that ORBs were not

4486 The Plant Cell

observed in the cryofixed wild-type anthers (Figures 4A and 4B).By contrast, ORBs were observed in cryofixed acos5 mutantanthers, located on all faces of tapetal cells (Figure 4C), as well asalong the surface of some free microspores in place of exine (Fig-ures 4D and 4F). The ORBs were electron translucent with a smallfibrillar electron-dense core (Figure 4E). The ORBs appeared to beembedded in the tapetum, outside the plasma membrane, on theinner periclinal (locule) side (Figures 4C and 4E). To test if thecryofixation and resin embedding for TEM had altered the ORBs ortheir locations, unfixed acos5 anthers were frozen, cryoplaned,and examined with cryoscanning electron microscopy, whichconfirmed the abundance of these bodies and their locationextracellular to tapetal cells (Figure 4G). ORBs were also observedaround some microspores in cryoplaned locules, while other mi-crospores lacked them (Figure 4H).As ACOS5 is an acyl-CoA synthetase postulated to act at the

start of the polyketide pathway (Quilichini et al., 2014b), mutantsin genes encoding the other polyketide pathway enzymes, suchas pksa pksb and tkpr1, were also examined with TEM. ORBs ofsimilar electron density and location to those in acos5 mutantswere observed (Supplemental Figure 6). The presence of extra-cellular ORBs in acos5, pksa pksb, and tkpr1 mutants suggestedthat tapetal cells produce fluorescent compounds in addition tothose generated by the polyketide pathway, but of unknowncomposition. If ABCG26 is required for the transport of theseunknown ORB components, then double or triple mutants shouldlack ORBs, but the abcg26 acos5 double mutant (Figures 2C and2D), abcg26 pksa pksb triple mutant (Figures 2E and 2F), andabcg26 tkpr1 double mutant (Figures 2G and 2H) all containORBs. This indicates that the ORB components can be exportedfrom tapetal cells in the absence of a functional ABCG26 trans-porter. The accumulation of these compounds at extracellularsurfaces of the tapetum and some microspore surfaces wasobserved only when polyketide product synthesis was directlyaffected, since the fluorescent bodies were not observed in theabcg26 mutant, where transporter loss of function presumablyimpacts polyketide product transport but not synthesis (Figure 1).No differences in lipid accumulation could be seen in tapetum

lipids in acos5 and wild-type anthers stained with Nile Red(Supplemental Figure 7). The extracellular ORBs did not stain withNile Red (green color in Supplemental Figure 7), but their presencein acos5 locules was confirmed by their autofluorescence (redcolor in Supplemental Figure 7). As observed in abcg26 anthers,the cytoplasm of developing acos5 microspores stained readilywith Nile Red, but the walls, when present, only stained minimallycompared with the wild type (Supplemental Figure 7).

Role of Phenylpropanoid and Flavonoid BiosyntheticEnzymes in Sporopollenin Biosynthesis

Double mutant analyses of abcg26with key polyketide biosynthesismutants (acos5, pksa pksb, and tkpr1) supported the hypothesisthat ABCG26 exports a product of the polyketide biosyntheticpathway that is incorporated into sporopollenin. As loss of thepolyketide pathway led to the accumulation of fluorescent ORBs,we hypothesized that polyketides and ORB components are syn-thesized by two distinct pathways whose products come togetherin the locule to make sporopollenin. ORBs are autofluorescent and

Figure 2. Polyketide Synthesis Pathway Products Are Required to Formthe Tapetal Inclusions in abcg26.

2-P microscopy of live wild-type and mutant anthers shows short-wavelength autofluorescence in teal color.(A) and (B) Wild-type anthers showing bright emission from developingmicrospore walls (A), dim emission from tapetal cells (B), and lack ofautofluorescent tapetum vacuolar inclusions (arrows).(C) and (D) abcg26 acos5 anthers showing lack of tapetum vacuolarinclusions (arrows).(E) and (F) abcg26 pksa pksb triple mutant anthers showing lack of ta-petum vacuolar inclusions (arrows).(G) and (H) abcg26 tkpr1 double mutant anthers showing lack of tapetumvacuolar inclusions (arrows).Bars = 20 mm.

Polyketide Trafficking in Exine Formation 4487

lack staining with Nile Red, suggesting that the second pathwayproduces a phenolic product. Despite numerous reports that phe-nolic compounds such as hydroxycinnamic acids are present insporopollenin, there is little information regarding the enzymes thatsynthesize such components. Thus, we targeted mutants in genesencoding enzymes in the phenylpropanoid and flavonoid pathwaysfor anther analysis by 2-P microscopy. The anthers of mutantsin genes encoding enzymes from the sinapyl alcohol branch ofthe monolignol biosynthetic pathway (CYP84A1/FERULIC ACID5-HYDROXYLASE [F5H]/FAH1) and the flavonoid pathway (4CL3and CHALCONE SYNTHASE [CHS]/TT4) were analyzed by 2-Pmicroscopy. A mutant in SHT was also studied, as this enzyme is

required for the biosynthesis of HC spermidine conjugates that areconstituents of the pollen coat in Arabidopsis (Grienenberger et al.,2009).Previous studies demonstrated that overall autofluorescence

emission from mature pollen grains of sht, 4cl3, and tt4 mutantswas altered (Grienenberger et al., 2009; Dobritsa et al., 2011). Themature pollen grain autofluorescence would be the sum of bothsporopollenin and pollen coat components. Here, the ability of 2-Pmicroscopy to sample microspores during the early stages of pollencell wall development was exploited, as we could investigate theemission profiles of the intrinsic fluorescence of sporopollenin-richmicrospores following tetrad release and preceding tapetal celldeath, prior to the addition of pollen coat constituents. To captureboth the range of anther autofluorescence and the peak wave-length(s) where emission reached maximum intensity, we per-formed spectral emission scans on intact developing mutant andwild-type anthers using 720-nm 2-P excitation and a 400- to 580-nmemission-capture range. Intensity projections over the range ofwavelengths allowed us to capture emission spectra specific totraced microspore walls while avoiding emission from surround-ing sporophytic tissues (Supplemental Figure 8). The mean mi-crospore autofluorescent emission intensities and SE values forthe wild type, 4cl3-2, tt4-2, sht, and fah1-2 (n = 24 microspores)were normalized to the 460-nm peak, which reached its maximumintensity in wild-type samples, to correct for differences in emis-sion intensity between samples. The resulting relative intensitiesfrom microspore emissions for wild-type, sht, 4cl3-2, tt4-2, andfah1-2 microspores are summarized in Figure 5. The wild-typemicrospore emission spectrum consisted of one broad peak witha maximum intensity at 460 nm. As reported previously for maturepollen grains, 4cl3 and tt4 mutant microspore walls exhibitedbroader emission around the 460-nm peak relative to the wild type,with increased intensities below 460 nm and decreases above460 nm. The fah1 and sht mutants showed a different trend, withspectral emission shifts toward wavelengths longer than 460 nm,peaking at 510 and 540 nm, respectively. Most notably, theemission profile from sht microspore walls prior to tapetal celldeath differed dramatically from that of the wild type. The capturedsht emissions were predominantly in the longer wavelength (i.e.,the 495- to 550-nm channel), producing images with red-coloredmicrospores (Figures 6F to 6I), rather than the teal emission seenin the wild type (Figures 6A to 6D). These data indicate that theproducts synthesized by 4CL3, CHS/TT4, CYP84A1/FAH1, andSHT affect the fluorescence profile of the pollen exine, reflectingchanges in the composition of the pollen cell wall in the stageswhen it consists predominantly of sporopollenin. The dramaticfluorescence spectral profile differences between wild-type andsht microspore walls indicate that HC spermidines are present inthe microspore wall prior to tapetal cell death.

Polyketide Product-Dependent ORBs ContainHC Spermidines

If polyketides and ORB components are synthesized by twodistinct pathways whose products come together in the loculeto make sporopollenin, and ORBs represent unincorporatedphenolics, then generating double mutants of phenylpropanoidbiosynthetic genes and acos5 is predicted to abolish ORBs.

Figure 3. The acos5Mutant Exhibits ORBs, Not Observed in the Wild Type.

2-P microscopy of live wild-type and mutant anthers shows short-wavelength autofluorescence in teal.(A) and (B) Intact wild-type anthers showing microspore walls with brightautofluorescence (A) and tapetal cells with diffuse emission (B).(C) and (D) Intact acos5 anthers showing autofluorescent bodies aroundtapetal cells on all sides and some microspores (C). Abundant ORBs areseen along the tapetal cell outer periclinal faces (D).(E) and (F) Enlarged views of the boxed regions in (C) and (D), re-spectively. ORB circular puncta are indicated by arrows in (E).Lo, locule; Msp, microspore; Tp, tapetum. Bars in (A) to (D) = 20 mm;bars in (E) and (F) = 5 mm.

4488 The Plant Cell

When examined with 2-P microscopy, and in contrast withacos5 single mutants (Figures 6K to 6N), ORBs were completelyabsent in the sht acos5 double mutant anthers and were notfound surrounding tapetal cells or microspores (Figures 6P to6S). These results were supported by TEM analyses, whereabundant ORBs surrounded tapetal cells in acos5 mutants(Figure 6O) but were absent in sht acos5 double mutants (Figure6T). As negative controls for the TEM analyses, wild-type andsht single mutant anthers lacked ORBs (Figures 6E and 6J).These data suggest that the extracellular bodies observed inacos5 locules are composed of HC spermidines.By contrast, mutations in genes in the flavonoid and sinapyl

alcohol lignin monomer pathways did not affect the ORBs sur-rounding tapetal cells and some microspores in acos5. ORBswere observed in acos5 4cl3, acos5 tt4, and acos5 fah1-2 doublemutants (Supplemental Figure 9). These data suggest that theORBs are not composed of flavonoid-derived or sinapyl alcohol-based compounds. While 4CL activity would be expected to berequired for HC spermidine biosynthesis, 4CL is encoded by afamily of four genes in Arabidopsis (Ehlting et al., 1999; Hambergerand Hahlbrock, 2004), and residual 4CL activity from the expres-sion of one or more of these 4CL genes could provide sufficientactivity to support the HC spermidine accumulation observed in theacos5 4cl3 background.Taken together, these data indicate that the trafficking of HC

spermidines to developing microspore wall occurs prior to tapetumprogrammed cell death, in conjunction with the ABCG26-mediatedexport of polyketide pathway products. Since the putative HCspermidine-containing bodies accumulate when the polyketidepathway is disrupted, trafficking of the HC spermidines from thetapetum surface to developing microspores appears to require thepolyketides.

HC Spermidines Do Not Contribute to the Autofluorescenceof Vacuolar Inclusions in Tapetal Cells of the abcg26 Mutant

The presence of HC spermidine-containing ORBs in polyketidepathway mutants and in their respective double mutants with

Figure 4. Ultrastructure of ORBs in acos5 and abcg26 acos5 MutantAnther Locules.

(A) to (F), (I), and (J) TEM micrographs of tapetal and developing mi-crospore cell edges.(G) and (H) Cryoscanning electron micrographs of tapetal and de-veloping microspore cell edges.(A) and (B) Wild-type anthers.(C) to (F) acos5 anthers. The locule-facing plasma membrane of tapetalcells shows deep invaginations containing ORBs (the boxed region in [C]is enlarged in [E]). ORBs were found on all faces of acos5 tapetal cellsand appeared electron translucent, with some electron-dense debris (E).The bodies identified in acos5 were also found around some micro-spores (the boxed region in [D] is enlarged in [F]).(G) and (H) acos5 anthers showing tapetum and microspore cell edges.Arrows in (G) highlight the extracellular position of ORBs, between thetapetum and endothecium and between neighboring tapetal cells. Theaccumulation of acos5 ORBs around some microspores (arrow), but notothers, is clearly visualized in (H).(I) and (J) acos5 abcg26 double mutant anthers. ORBs were observed onall faces of the tapetum (I) and on some microspores (J).En, endothecium; Lo, locule; Msp, microspore; Tp, tapetum. Bars in (A)to (D) and (G) to (J) = 2 mm; bars in (E) and (F) = 0.5 mm.

Polyketide Trafficking in Exine Formation 4489

abcg26 indicated that HC spermidine trafficking from tapetalcells is independent of the ABCG26 transporter. However, theHC spermidine bodies did not accumulate in the abcg26 singlemutant, for reasons that are not clear. To examine the possibilitythat HC spermidines accumulate inside tapetal cells in the ab-sence of polyketide export, sht abcg26 double mutant antherswere examined in parallel to abcg26 anthers by 2-P microscopy.In sht abcg26 anthers, brightly fluorescent tapetum vacuolarinclusions similar to those observed in abcg26 were apparent(Figures 7A to 7H). The tapetum vacuolar inclusions observed insht abcg26 and abcg26 were indistinguishable by TEM (Figures7I and 7J). The persistence of tapetum vacuolar inclusions in shtabcg26 supports the interpretation that the vacuolar inclusionscontain products of the polyketide biosynthetic pathway in-volving ACOS5, PKSA PKSB, and TKPR1.

Incorporation of HC Spermidines in the Pollen Wall

While HC spermidines have been described previously as com-ponents of the pollen coat in Arabidopsis, their export from tapetalcells prior to programmed cell death suggested that they may beincorporated into sporopollenin during exine formation. To de-termine whether the spermidine conjugates are incorporated intothe sporopollenin biopolymer and thus contribute to exine in-tegrity, or remain more loosely attached as components of thepollen coat, the chemical recalcitrance of wild-type and shtmutantsporopollenin of mature pollen grains was assessed. To test thechemical integrity of sht pollen grains, pollen were subjected toacetolysis, a standard test for sporopollenin chemical integrity(representative images for three trials are shown in SupplementalFigure 10). However, these analyses failed to reveal any differ-ences between wild-type and sht pollen walls, which were both

insensitive to acetolysis, with 11.3% wild-type exines and 5.3%sht exines showing visible damage (n = 600 pollen grains pergenotype).

DISCUSSION

In this study, we provide insights into the nature of sporopolleninand pollen wall exine formation. The genetic link found betweenthe ABCG26 transporter and the polyketide pathway involved insporopollenin biosynthesis provides evidence that the substratetransported by ABCG26 is a polyketide product. Furthermore,our analyses showed that HC spermidine conjugates known tobe present in the Arabidopsis pollen coat are trafficked fromtapetal cells prior to programmed cell death in an ABCG26-independent fashion. Although polyketide components are ableto assemble in the absence of HC spermidines, spermidineconjugates required the polyketide component of sporopolleninfor normal deposition on microspores. One interpretation of thesedata is that copolymers of polyketides and HC spermidines trafficfrom the tapetum to the microspore wall, prior to deposition intothe exine of developing microspores. In addition to revealingmechanisms of pollen exine traffic not understood previously, thenovel 2-P microscopy approach employed here has opened thedoor to the study of intact anthers and provides a useful tool forfuture studies.

In Planta Characterization of the Putative Substrateof ABCG26

Previous studies that relied on the examination of fixed and sec-tioned tapetal cells were unable to identify accumulations inabcg26mutant tapeta that could be indicative of putative ABCG26

Figure 5. Spectral Intrinsic Fluorescence Emission Profiles for Wild-Type, 4cl3, tt4, sht, and fah1 Mutant Microspores within Developing Anthers andPrior to Tapetal Cell Death.

Intrinsic fluorescence of microspore walls was collected using 720-nm 2-P excitation, providing effective excitation in the near UV range. Spectralemission scans in the 400- to 580-nm interval were obtained with a 10-nm step size and normalized to the 460-nm peak emission intensity. For eachgenotype, four microspores per locule were traced from six anthers, and mean 6 SE emission intensity at each 10-nm wavelength was plotted.

4490 The Plant Cell

substrates. By contrast, 2-P microscopy identified inclusions inabcg26 tapeta in the form of large autofluorescent vacuoles. Thisinformation facilitated a targeted genetic approach to determinethe biosynthetic origin of these inclusions as polyketide products.Based on the first appearance of these vacuolar inclusions inabcg26 tapetal cells immediately following the tetrad stage, fol-lowed by their persistence over the course of free microsporedevelopment, the ABCG26 transporter appears to be active overthis developmental period, consistent with the timing of its ex-pression in tapetal cells (Quilichini et al., 2010).

Analysis of lipidic metabolite accumulation in live tapetal cells,using Nile Red, revealed similar abundances of lipids in wild-typeand abcg26 cells. Thus, the compounds accumulating in abcg26tapetal cells differ dramatically in nature from the putative cuticularsubstrates that accumulate in abcg11 and abcg12 mutants, twohalf-size ABCG transporters in the same subfamily (Pighin et al.,2004; Bird et al., 2007). Accumulations in abcg11 and abcg12epidermal cells stain with Nile Red, appear lamellar, and are as-sociated with the endoplasmic reticulum (Pighin et al., 2004; Birdet al., 2007; McFarlane et al., 2010). Therefore, the ABCG26

Figure 6. HC Spermidines Make up ORBs in acos5.

(A) to (D), (F) to (I), (K) to (N), and (P) to (S) 2-P microscopy of live wild-type and mutant anthers.(E), (J), (O), and (T) TEM images of cryofixed anthers.(A) to (D) Wild-type microspores and tapetal cells. Microspore wall autofluorescent emission is preferentially in short wavelengths (420 to 460 nm;colored teal).(E) Wild-type tapetal cell edges show a lack of ORBs.(F) to (I) sht microspores and tapetal cells. Microspore autofluorescent emission is predominantly in longer wavelengths (495 to 540 nm; colored red)rather than in short wavelengths (teal).(J) sht tapetal cell edges show a lack of ORBs.(K) to (N) acos5microspores and tapetal cells. Brightly autofluorescent ORBs are prominent surrounding tapetal cells, emitting at 420 to 460 nm (colored teal).(O) acos5 tapetal cell edge showing extracellular ORBs.(P) to (S) sht acos5 microspores and tapetal cells. No autofluorescent ORBs were observed surrounding tapetal cells or microspores.(T) sht acos5 tapetal cell edges show a lack of ORBs.Tp, tapetum. Magnified columns show enlarged views of the boxed regions. Bars in (A), (C), (F), (H), (K), (M), (P), and (R) = 20 mm; bars in (B), (D), (G),(I), (L), (N), (Q), and (S) = 5 mm; bars in (E), (J), (O), and (T) = 2 mm.

Polyketide Trafficking in Exine Formation 4491

substrate is likely a unique polyketide-derived, autofluorescentcompound translocated from tapetal cells to the microspores inthe locule, distinct from the apparent lipidic cargo of ABCG11 andABCG12.

The Polyketide Biosynthetic Pathway and Transport byABCG26 Are Linked

Only a handful of enzymes required for sporopollenin synthesisin Arabidopsis have been characterized biochemically. Amongthese, ACOS5, PKSA, PKSB, and TKPR1 are proposed tofunction sequentially in the production of tetraketide a-pyrones(Grienenberger et al., 2010; Quilichini et al., 2014b). The strongphenotypes of acos5, pksa pksb, and tkpr1 mutants that, likeabcg26, are male sterile and lack sporopollenin deposition, aswell as the overlapping spatial and temporal expression patternsshared by the biosynthetic genes and ABCG26, suggested thatpolyketides produced by these enzymes could be exported byABCG26 (Grienenberger et al., 2010; Quilichini et al., 2010). Wedirectly tested this hypothesis by examining the phenotypes ofdouble and triple mutant combinations affecting each of theenzymes in the pathway in the abcg26 mutant background. 2-Pimaging and TEM analyses of the abcg26 acos5, abcg26 pksapksb, and abcg26 tkpr1 mutants revealed phenotypes identicalto the acos5, pksa pksb, and tkpr1mutants, including the lack ofautofluorescent vacuolar inclusions in tapetal cells. The sup-pression of the abcg26 tapetum vacuolar inclusion phenotype ineach of these mutant backgrounds provides direct evidence thatthe polyketide biosynthetic pathway is required for these ABCG26-dependent tapetal cell inclusions. Thus, the simplest interpretationis that tetraketide a-pyrone end products of the polyketide path-ways, or their derivatives, are in fact the ABCG26 substrates.Transport assays will be required to define the ABCG26 substratespecificity precisely, and the lack of vacuolar inclusions in all poly-ketide mutants independently of their position in the pathway maysuggest that autofluorescent inclusions in abcg26 are caused by theaccumulation of more derived compounds than the TKPR1-produced tetraketide a-pyrones themselves. In Figure 8A, we pro-pose an updated model for pollen exine formation based on thedata presented here, illustrating the function of the ABCG26 trans-port protein in the export of polyketide pathway products from thetapetum. The key mutant phenotypes observed from our geneticanalyses, which inform the model, are illustrated in Figures 8B to 8G.The finding that tapetum inclusions in abcg26 likely represent

products of the polyketide pathway provides a better understanding

Figure 7. abcg26 and sht abcg26 Tapetum Vacuolar Inclusions Are In-distinguishable by 2-P Microscopy and TEM.

(A), (C), (E), and (G) 2-P microscopy of live abcg26 mutant anthers.(B), (D), (F), and (H) 2-P microscopy of live sht abcg26 mutant anthers.

(A) to (D) Bright intrinsic fluorescence was emitted at the locule pe-riphery, in tapetum and middle layer cells. (C) and (D) show enlarge-ments of the boxed regions in (A) and (B), respectively.(E) to (H) abcg26 and sht abcg26 tapetal cells contain autofluorescentvacuoles with numerous brightly emitting punctae. (G) and (H) showenlargements of the boxed regions in (E) and (F), respectively,(I) and (J) TEM of abcg26 (I) and sht abcg26 (J) tapetal cell vacuolesshows electron-dense circular cores, which varied in number and sizebetween samples. No consistent differences were observed between theinclusions in abcg26 and sht abcg26 anthers.Bars in (A), (B), (E), and (F) = 20 mm; bars in (C), (D), (G), and (H) = 5 mm;bars in (I) and (J) = 1 mm.

4492 The Plant Cell

of the roles played by polyketide pathway product(s) in exine for-mation in planta. The bright fluorescence of the abcg26 tapetum-trapped metabolites suggests the presence of aromatic rings orconjugated p bonds in the metabolites. While the proposedproduct of the sporopollenin polyketide metabolon is a tetraketidea-pyrone, there is limited literature on the intrinsic fluorescentproperties of a-pyrone rings. It is also possible that the tetraketideproduct of this polyketide pathway may be further modified orcyclized prior to export from the tapetum (Figure 8A). Polyketidecyclization by type III polyketide synthases, such as PKSA andPKSB, is known to vary, encompassing heterocyclic lactoneformation and intramolecular condensations either by aldol orClaisen carbon-carbon bond formation, or to be absent alto-gether (Austin and Noel, 2003; Funa et al., 2006; Cook et al.,2010). InCannabis, the cooperative activity of two proteins, a typeIII PKS and an accessory protein called olivetolic acid cyclase, isrequired for the correct cyclization of polyketides in the pro-duction of a key intermediate in tetrahydrocannabinolic acidsynthesis (Gagne et al., 2012). Without olivetolic acid cyclase, thepolyketide enzyme forms olivetol and a-pyrone by-products. Theoccurrence and function of a-pyrones in plants are minimallyunderstood, and some reports have suggested that a-pyronesare derailment by-products of in vitro assays (Akiyama et al.,1999; Yamaguchi et al., 1999; Abe et al., 2004). However, novelpathways capable of pyrone formation in plants have been de-scribed recently, and their presence in the sporopollenin bio-polymer is plausible (Eckermann et al., 1998; Weng et al., 2012).Thus, the exact chemical nature of the autofluorescent polyketidepathway-dependent product that appears to be the ABCG26substrate for export into the locule remains to be determined.

ORBs in Polyketide Mutants Provide Insight into theTrafficking of Novel Exine Components

The suppression of ABCG26-dependent tapetum vacuolar in-clusions by polyketide biosynthetic mutants provides evidencefor a functional link between the synthesis and export of poly-ketides, both of which are required for sporopollenin formation(Figures 8B, 8C, and 8E). However, a second and more puzzlingphenotype was observed in mutants blocked in polyketidebiosynthesis. Locule-localized autofluorescent ORBs were ob-served in acos5, pksa pksb, tkpr1, and their respective doublemutants in the abcg26 background by 2-P microscopy (Figures2 and 3; Supplemental Figure 5). These had obvious similarity tothe orbicules observed in the locules of some species in asso-ciation with exine deposition (Huysmans et al., 1998; Wanget al., 2003). The presence of ORBs in the absence of polyketidebiosynthesis suggested the existence of a second pathway thatrequires the polyketide component of sporopollenin in order tobe assembled into the exine and that these metabolites areexported into the locule independently of ABCG26.

HC Spermidines Are Exported from Tapetal Cells Prior toProgrammed Cell Death and Are Incorporated into Exine

The ORBs observed in polyketide biosynthetic pathway mutantsprovide information that significantly extends current understandingof exine assembly. The autofluorescent nature of ORBs supports

the presence of aromatic rings, and their exclusion of Nile Red stainsuggests a nonhydrophobic composition, while their dependenceon a functional polyketide pathway suggests interaction with thesemetabolites. In the acos5 mutant background, the sht mutant,blocked in HC spermidine synthesis, suppressed ORB formation(Figure 6 and illustrated in Figure 8F), but mutants blocking otherphenylpropanoid branch pathways leading to flavonoid and sina-poyl alcohol biosynthesis had no effect.While sht mutants produce fertile, morphologically normal pol-

len (Grienenberger et al., 2009), the exine of developing sht mi-crospores exhibited significantly shifted emission spectra relativeto wild-type microspores prior to tapetum programmed cell deathbased on 2-P microscopy, indicating a change in exine compo-sition in the sht mutant early in microspore wall development(Figures 6F and 6G and illustrated in Figure 8D). These data in-dicate that HC spermidines, previously characterized componentsof the pollen coat, are not contributed solely after tapetum pro-grammed cell death but are exported from tapetal cells with similartiming to the sporopollenin components contributed by the poly-ketide pathway. This provides evidence for the premortem traf-ficking of pollen coat components from the tapetum and suggeststhat, because HC spermidines are present in the locule at the timeof sporopollenin assembly, these compounds could be incorpor-ated into the growing polymer.The dependence of HC spermidine accumulation in ORBs upon

polyketide biosynthesis raises intriguing questions regarding thefunctional link between these two apparent exine components.The polyketide component of sporopollenin is capable of as-sembly in the absence of HC spermidines, as seen from the shtmutant, while the HC spermidines require polyketide synthesis toassemble on the pollen wall. One plausible interpretation of thesedata is that polyketides form an exine framework by self-assem-bly, to which HC spermidines may adhere during the phase ofrapid exine development on microspores released from tetrads.Since HC spermidines coalesce into ORBs in the absence ofpolyketide biosynthesis, cotrafficking and potential physical in-teraction of HC spermidines with polyketides upon their releaseinto the locule is an attractive model (Figure 8A). The presence ofHC spermidine-containing ORBs around some microspores, andtheir absence around others, is puzzling and suggests that un-known anchoring or targeting mechanisms may also be in place.Finally, an alternative explanation for ORB formation in polyketidemutants is that the metabolic block in the polyketide pathwayintroduced by acos5 and other mutants in the pathway leads toincreased metabolic flux into the HC spermidine pathway (Wenget al., 2010).Since HC spermidine-containing ORBs still formed when

polyketide biosynthetic mutations were placed in the abcg26mutant background, their export from tapetal cells appears to beABCG26 independent (Figure 8E). Although the mechanism forHC spermidine export from tapetal cells is unknown, it is in-teresting that these bodies were observed on all tapetum facesin polyketide mutants, suggesting that the tapetal cells may notbe polarized in their secretion to the locule.At present, the lack of HC spermidine-containing ORBs in the

abcg26 single mutant is puzzling. According to our model, abcg26anthers should lack polyketide accumulation in locules, similar toacos5, acos5 abcg26, and other polyketide mutants in wild-type or

Polyketide Trafficking in Exine Formation 4493

abcg26 backgrounds in which ORBs are prominent. However, ifonly the transport of polyketides, and not their biosynthesis, isaffected in the abcg26 transporter mutant, then residual transportof some polyketide product into the locule could explain these re-sults. Residual transport could occur from the presence of several

other ABCG transporters, whose expression in the tapetum hasbeen demonstrated (Ariizumi and Toriyama, 2011). Alternatively,the export of HC spermidines from intact tapetal cells could beinhibited by the accumulation of polyketides in tapetal cellsbrought about by their blocked transport. This, in turn, would

Figure 8. A Model for Outer Pollen Wall Biosynthesis, Transport, and Assembly in Arabidopsis.

(A) Two tapetum-localized biosynthetic pathways, defined herein as the polyketide pathway and the HC spermidine pathway, are presented. In thepolyketide pathway (colored green), a sequence of biosynthetic reactions defined by ACOS5, PKSA, PKSB, TKPR1, and possibly two cytochromeP450s, CYP703A2 and CYP704B1, produces hydroxylated tetraketide products from plastid-produced fatty acids (for detailed sporopollenin bio-synthetic models, see Grienenberger et al. [2010] and Quilichini et al. [2014b]). As these products form a-pyrone rings in vitro, the possibility thata currently unknown cyclase enzyme may function in phenolic ring formation is noted. The export of polyketide pathway products requires the ABCG26half-size transport protein, which may homodimerize or heterodimerize for export. In the pathway synthesizing HC spermidines (colored purple), SHTconjugates HC-CoA esters to spermidine. Independent from the ABCG26 transporter, the HC spermidine products exit the tapetum by an unknownmechanism prior to programmed cell death. In the anther locule, polyketide and HC spermidine pathway products may interact or traffic as copolymersprior to deposition into the pollen exine. ER, endoplasmic reticulum; G, Golgi apparatus; M, mitochondrion; N, nucleus; P, plastid; V, vacuole.(B) to (G) Select Arabidopsis mutants that inform the model shown in (A).(B) The abcg26mutant microspores, lacking exine formation, begin to abort, and tapetal cells accumulate autofluorescent vacuolar inclusions, shown in blue.(C) The acos5 mutant affecting the polyketide pathway exhibits autofluorescent ORBs, shown in blue, around select microspores and tapetal cells.(D) The sht mutant affecting the HC spermidine pathway exhibits assembled exine walls, colored red, that emit at longer wavelengths than in the wildtype.(E) The abcg26 acos5 double mutant lacks the tapetum vacuolar inclusions observed in abcg26 single mutants (B) and contains ORBs as in acos5single mutants (C).(F) The sht acos5 double mutant lacks tapetum vacuolar inclusions and ORBs.(G) The sht abcg26 double mutant, as in the abcg26 single mutant (B), exhibits tapetum vacuolar inclusions.

4494 The Plant Cell

suggest an abnormal interaction between polyketides and HCspermidines within tapetal cells or the existence of a feedbackmechanism repressing HC spermidine biosynthesis upon ab-normal polyketide accumulation in tapetal cells.

Finally, the functions of HC spermidines that appear to bedeposited into the exine in parallel with polyketide a-pyronesremain to be established. Mutants lacking SHT function are fertile(Grienenberger et al., 2009), and we found no difference in thesensitivity of sht versus wild-type pollen grains to acetolysis(Supplemental Figure 10), suggesting that the robustness of pollenwall exine is not compromised in the absence of this component.However, it is possible that HC spermidine components play moresubtle roles in resistance to environmental stresses such asdesiccation or UV irradiance.

In summary, our results from live-cell imaging of Arabidopsisanthers reveal that the cargo transported by ABCG26 includespolyketide pathway components of sporopollenin and that HCspermidines are transported from tapetal cells into the antherlocule prior to tapetum programmed cell death. A complex, and asof yet poorly understood, interplay between polyketide productsand HC spermidine biosynthesis appears to govern their depositionon the microspore surface and will be the basis for future studies.

METHODS

Plant Growth

Arabidopsis thaliana seeds were germinated on Murashige and Skoogmedium plates, pH 5.7. Seeds were imbibed in the dark at 4°C, grownunder continuous light for 7 to 10 d, and then transplanted to soil(SunshineMix 4; Sungrow Horticulture). Seedlings were grown to maturityunder long-day conditions (16-h-light/8-h-dark cycles) at 20°C.

2-P Laser Scanning Microscopy Analysis

For stamen dissection, unopened flower buds were excised from plantsusing anOlympusSZX10 stereomicroscope. Budsmeasuring 0.7 to 1.2mmwere used for stamen collection, providing anthers in the free microsporestage. Stamens from one bud were immersed in distilled water for intrinsicfluorescence analysis or Nile Red staining solution for lipid analysis. Airbubbles were removed by gentle manipulation, and samples were coveredwith a 1.5-mm cover slip. Stamens were examined by 2-P microscopyimmediately following dissection (and staining, when applicable) using theOlympus Multi-photon Laser Scanning Microscope FV1000MPE and theOlympus XLPLN 25X WMP dedicated objective.

Anther intrinsic fluorescence was examined using 720-nm excitationand the BFP/GFP/RFP/DsRed filter cube for four-channel imaging ofstamens. Emission was captured in two channels, RXD1 (420 to 460 nm,colored teal) and RXD2 (495 to 540 nm, colored red).

Nile Red dye (9-diethylamino-5H-benzo[a]phenoxazine-5-one) stocksolution was prepared in acetone (0.1 mg/mL), and fresh working sol-utions were made in sterile 25% glycerol (1 mg/mL) (Greenspan et al.,1985). Microscope settings optimal for intracellular Nile Red fluorescenceemission detection were determined by comparing the fluorescenceemission profiles for wild-type anthers in 25% glycerol and 25% glycerolwith 1 mg/mL Nile Red (Supplemental Figure 2). An 810-nm excitation andthe CFP/YFP/RFP/DsRed filter cube (Olympus) were used for two-channelimaging, capturing emission from RXD3 (380 to 560 nm, colored red) forintrinsic fluorescence and RXD4 (575 to 630 nm, colored green) for NileRed. Unstained anthers fluoresced in the RXD3 channel, with only traces offluorescence emission in theRXD4 channel. Wild-type anthers immersed in

Nile Red staining solution emitted Nile Red fluorescence initially along theanther cuticle, and microspore wall staining became visible after 20 to30 min. Olympus FV10-ASW version 03.01 and Volocity version 6.1.1software packages were used for image analyses.

Spectral Analysis (Lambda Scans)

After orienting samples with the UV intrinsic fluorescence settings de-scribed above, spectral analyses (lambda scans) were performed with15% laser power (with the exception of tt4-2 mutants) at 720-nm exci-tation, with 800 high voltage on the photomultiplier tube (PMT) detector,8 s/mm pixel dwell time, and a 256 3 256 pixel density, with RDM690,VBF, and PMT1 detector selections in place. Emission was collectedbetween 400 and 600 nm, with a 10-nm bandwidth and step size.Samples that burst during spectral analysis were not included. Followingspectral scans, intensity projection over the lambda axis was used totrace microspore wall fluorescence (n = 24 pollen grains per trial), andemission from traced regions of interest was plotted usingOlympus FV10-ASW version 03.01.

TEM Analysis

Unopened flower buds with sepals and petals excised were submerged in0.2 M sucrose in uncoated copper type B sample holders (Ted Pella).Samples were frozen at high pressure with a Leica EM HPM 100 high-pressure freezer, followed by freeze substitution and embedding in epoxyresin as described by Quilichini et al. (2014a).

Cryoscanning Electron Microscopy Analysis

Dissected anthers from unopened flower buds (measuring 0.7 to 1.2 mmlong) were submerged in 0.2 M sucrose between two specimen carriers(Dia. 3.03 0.8mmwith cylinder-shaped indentation; LeicaMicrosystems)and frozen under high pressure (EM HPM100; Leica Microsystems). Theupper specimen carrier was removed under liquid nitrogen to exposeanthers after freezing. The remaining sample-filled specimen carrier wasclamped to a cold table (also called a specimen holder for freeze frac-turing; Leica Microsystems) and transferred under liquid nitrogen toa cryomicrotome (Leica EM UC7 microtome; precooled to 2140°C). Toexpose one or more anther locule(s), cryoplaning was performed witha diamond knife (Cryo 35°; Diatome), 200-nm trimming thickness, and anantistatic line on. Each cryoplaned sample was transferred under vacuum(using the Leica VCT100) to a precooled scanning cryoelectron micro-scope (2140°C; S-4700 Field Emission scanning electron microscope;Hitachi). Sublimation at 2105°C for 5 min was applied to remove surfaceice, followed by cooling to2130°C for subsequent imaging. Imaging wasperformed at 2.0 kVwith an 11-mmworking distance for two independentlyprepared, unstained acos5 anther samples.

Acetolysis

A pellet of pollen from 10 open flowers was washed with glacial aceticacid, placed in acetolysis solution (nine parts acetic anhydride to one partconcentrated sulfuric acid, adding the acid slowly), sealed, and boiled for10 min in a fume hood. Remaining exines were washed in glacial aceticacid, then three times with water, and imaged with a light microscope(Leica DMR microscope). Three trials were performed, with 200 exinescounted per trial.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis GenomeInitiative database under the following accession numbers: At3g13220(ABCG26), At1g62940 (ACOS5), At1g02050 (PKSA), At4g34850 (PKSB),

Polyketide Trafficking in Exine Formation 4495

At4g35420 (TKPR1), At4g36220 (FAH1), At1g65060 (4CL3), At5g13930(TT4), and At2g19070 (SHT).

Supplemental Data

The following materials are available in the online version of this article.

Supplemental Figure 1. Variation in the Severity of the abcg26 AntherPhenotype.

Supplemental Figure 2. Controls for the Timing of Nile Red Stainingof Anthers.

Supplemental Figure 3. abcg26 Tapetum Vacuole Cores Are Hydro-phobic in Nature.

Supplemental Figure 4. Vacuolar Inclusions Are Present in abcg264cl3 and abcg26 tt4 Mutant Tapetal Cells.

Supplemental Figure 5. cyp703a2, pksa pksb, and tkpr1 MutantsExhibit Tapetum-Associated Autofluorescent Orbicule-Like Bodies(ORBs), Not Observed in the Wild Type.

Supplemental Figure 6. Extracellular ORBs Are Observed in pksapksb and tkpr1 Anther Locules.

Supplemental Figure 7. Extracellular Autofluorescent ORBs in acos5Anther Locules Do Not Stain with Nile Red.

Supplemental Figure 8.Microspore Emission from Traced Regions ofInterest in Developing Anthers.

Supplemental Figure 9. The acos5 4cl3, acos5 tt4, and acos5 fah1Double Mutants Exhibit the Same ORB Phenotype as the acos5 SingleMutant

Supplemental Figure 10. The Chemical Recalcitrance of the Sporo-pollenin Polymer in Wild-Type and sht Mutant Pollen Is Indistinguish-able.

Supplemental Movie 1. Z-Stack of Two-Photon Images of a Wild-Type Col-0 Anther.

Supplemental Movie 2. Z-Stack of Two-Photon Images of an abcg26Anther.

ACKNOWLEDGMENTS

We thank TAIR (Carnegie Institute, Stanford, CA) and the ABRC (OhioState University) for providing Arabidopsis T-DNA insertion mutants. Wethank the University of British Columbia BioImaging Facility staff fortechnical assistance and gratefully acknowledge laboratory assistantsHans Liu, Samantha Kang, and Ada Roman. We are grateful for manuscriptsuggestions from Ljerka Kunst, Reinhard Jetter, Heather McFarlane, andEtienne Grienenberger. This work was supported by the Canadian NaturalSciences and Engineering Research Council (NSERC) Discovery Grants toC.J.D. and A.L.S., an NSERC-PGSD award to T.D.Q., and the NSERCCREATE Working on Walls program.

AUTHOR CONTRIBUTIONS

T.D.Q., A.L.S., and C.J.D. participated in research design. T.D.Q.conducted experiments. T.D.Q., A.L.S., and C.J.D. carried out dataanalysis. T.D.Q., A.L.S., and C.J.D. wrote the article.

Received August 1, 2014; revised October 14, 2014; accepted November6, 2014; published November 21, 2014.

REFERENCES

Aarts, M.G.M., Hodge, R., Kalantidis, K., Florack, D., Wilson, Z.A.,Mulligan, B.J., Stiekema, W.J., Scott, R., and Pereira, A. (1997).The Arabidopsis MALE STERILITY 2 protein shares similarity withreductases in elongation/condensation complexes. Plant J. 12:615–623.

Abe, I., Watanabe, T., and Noguchi, H. (2004). Enzymatic formationof long-chain polyketide pyrones by plant type III polyketide syn-thases. Phytochemistry 65: 2447–2453.

Ahlers, F., Lambert, J., and Wiermann, R. (2003). Acetylation andsilylation of piperidine solubilized sporopollenin from pollen ofTypha angustifolia L. Z. Naturforsch. C 58: 807–811.

Ahlers, F., Thom, I., Lambert, J., Kuckuk, R., and Wiermann, R.(1999). 1H NMR analysis of sporopollenin from Typha angustifolia.Phytochemistry 50: 1095–1098.

Akiyama, T., Shibuya, M., Liu, H.-M., and Ebizuka, Y. (1999). p-Coumaroyltriacetic acid synthase, a new homologue of chalconesynthase, from Hydrangea macrophylla var. thunbergii. Eur. J. Bio-chem. 263: 834–839.

Ariizumi, T., and Toriyama, K. (2011). Genetic regulation of sporo-pollenin synthesis and pollen exine development. Annu. Rev. PlantBiol. 62: 437–460.

Austin, M.B., and Noel, J.P. (2003). The chalcone synthase super-family of type III polyketide synthases. Nat. Prod. Rep. 20: 79–110.

Bird, D., Beisson, F., Brigham, A., Shin, J., Greer, S., Jetter, R.,Kunst, L., Wu, X., Yephremov, A., and Samuels, L. (2007). Char-acterization of Arabidopsis ABCG11/WBC11, an ATP binding cas-sette (ABC) transporter that is required for cuticular lipid secretion.Plant J. 52: 485–498.

Bullock, G.R. (2011). The current status of fixation for electron mi-croscopy: A review. J. Microsc. 133: 1–15.

Buschmann, C., Langsdorf, G., and Lichtenthaler, H.K. (2000).Imaging of the blue, green, and red fluorescence emission of plants:An overview. Photosynthetica 38: 483–491.

Cerovic, Z.G., Samson, G., Morales, F., Tremblay, N., and Moya, I.(1999). Ultraviolet-induced fluorescence for plant monitoring: Presentstate and prospects. Agronomie 19: 543–578.

Chen, W., Yu, X.H., Zhang, K., Shi, J., De Oliveira, S., Schreiber, L.,Shanklin, J., and Zhang, D. (2011). Male Sterile2 encodes a plas-tid-localized fatty acyl carrier protein reductase required for pollenexine development in Arabidopsis. Plant Physiol. 157: 842–853.

Choi, H., Jin, J.-Y., Choi, S., Hwang, J.-U., Kim, Y.-Y., Suh, M.C.,and Lee, Y. (2011). An ABCG/WBC-type ABC transporter is es-sential for transport of sporopollenin precursors for exine formationin developing pollen. Plant J. 65: 181–193.

Cook, D., Rimando, A.M., Clemente, T.E., Schröder, J., Dayan, F.E.,Nanayakkara, N.P.D., Pan, Z., Noonan, B.P., Fishbein, M., Abe, I.,Duke, S.O., and Baerson, S.R. (2010). Alkylresorcinol synthases ex-pressed in Sorghum bicolor root hairs play an essential role in the bio-synthesis of the allelopathic benzoquinone sorgoleone. Plant Cell 22:867–887.

Descolas-Gros, C., and Schölzel, C. (2007). Stable isotope ratios ofcarbon and nitrogen in pollen grains in order to characterize plantfunctional groups and photosynthetic pathway types. New Phytol.176: 390–401.

Dobritsa, A.A., Geanconteri, A., Shrestha, J., Carlson, A., Kooyers,N., Coerper, D., Urbanczyk-Wochniak, E., Bench, B.J., Sumner,L.W., Swanson, R., and Preuss, D. (2011). A large-scale geneticscreen in Arabidopsis to identify genes involved in pollen exineproduction. Plant Physiol. 157: 947–970.

Dobritsa, A.A., Lei, Z., Nishikawa, S., Urbanczyk-Wochniak, E.,Huhman, D.V., Preuss, D., and Sumner, L.W. (2010). LAP5 andLAP6 encode anther-specific proteins with similarity to chalcone

4496 The Plant Cell

synthase essential for pollen exine development in Arabidopsis.Plant Physiol. 153: 937–955.

Dobritsa, A.A., Shrestha, J., Morant, M., Pinot, F., Matsuno, M.,Swanson, R., Møller, B.L., and Preuss, D. (2009). CYP704B1 isa long-chain fatty acid omega-hydroxylase essential for sporopolleninsynthesis in pollen of Arabidopsis. Plant Physiol. 151: 574–589.

Dou, X.-Y., Yang, K.-Z., Zhang, Y., Wang, W., Liu, X.-L., Chen, L.-Q.,Zhang, X.-Q., and Ye, D. (2011). WBC27, an adenosine tri-phosphate-binding cassette protein, controls pollen wall formation and patterningin Arabidopsis. J. Integr. Plant Biol. 53: 74–88.

Eckermann, S., Schröder, G., Schmidt, J., Strack, D., Edrada, R.A.,Helariutta, Y., Elomaa, P., Kotilainen, M., Kilpeläinen, I., and Proksch,P. (1998). New pathway to polyketides in plants. Nature 396: 387–390.

Ehlting, J., Büttner, D., Wang, Q., Douglas, C.J., Somssich, I.E.,and Kombrink, E. (1999). Three 4-coumarate:coenzyme A ligasesin Arabidopsis thaliana represent two evolutionarily divergentclasses in angiosperms. Plant J. 19: 9–20.

Funa, N., Ozawa, H., Hirata, A., and Horinouchi, S. (2006). Phenoliclipid synthesis by type III polyketide synthases is essential for cystformation in Azotobacter vinelandii. Proc. Natl. Acad. Sci. USA 103:6356–6361.

Gagne, S.J., Stout, J.M., Liu, E., Boubakir, Z., Clark, S.M., andPage, J.E. (2012). Identification of olivetolic acid cyclase fromCannabis sativa reveals a unique catalytic route to plant poly-ketides. Proc. Natl. Acad. Sci. USA 109: 12811–12816.

Greenspan, P., Mayer, E.P., and Fowler, S.D. (1985). Nile red: a se-lective fluorescent stain for intracellular lipid droplets. J. Cell Biol.100: 965–973.

Grienenberger, E., Besseau, S., Geoffroy, P., Debayle, D., Heintz,D., Lapierre, C., Pollet, B., Heitz, T., and Legrand, M. (2009). ABAHD acyltransferase is expressed in the tapetum of Arabidopsisanthers and is involved in the synthesis of hydroxycinnamoylspermidines. Plant J. 58: 246–259.

Grienenberger, E., Kim, S.S., Lallemand, B., Geoffroy, P., Heintz,D., de Azevedo Souza, C., Heitz, T., Douglas, C.J., and Legrand,M. (2010). Analysis of TETRAKETIDE a-PYRONE REDUCTASEfunction in Arabidopsis thaliana reveals a previously unknown, butconserved, biochemical pathway in sporopollenin monomer bio-synthesis. Plant Cell 22: 4067–4083.

Gubatz, S., Rittscher, M., Meuter, A., Nagler, A., andWiermann, R. (1993).Tracer experiments on sporopollenin biosynthesis. Grana 32: 12–17.

Guilford, W.J., Schneider, D.M., Labovitz, J., and Opella, S.J.(1988). High resolution solid state 13C NMR spectroscopy of spor-opollenins from different plant taxa. Plant Physiol. 86: 134–136.

Hamberger, B., and Hahlbrock, K. (2004). The 4-coumarate:CoAligase gene family in Arabidopsis thaliana comprises one rare, sinapate-activating and three commonly occurring isoenzymes. Proc. Natl. Acad.Sci. USA 101: 2209–2214.

Hsieh, K., and Huang, A.H. (2007). Tapetosomes in Brassica tapetumaccumulate endoplasmic reticulum-derived flavonoids and alkanesfor delivery to the pollen surface. Plant Cell 19: 582–596.

Huysmans, S., El-Ghazaly, G., and Smets, E. (1998). Orbicules inangiosperms: morphology, function, distribution, and relation withtapetum types. Bot. Rev. 64: 240–272.

Kim, S.S., et al. (2010). LAP6/POLYKETIDE SYNTHASE A and LAP5/POLYKETIDE SYNTHASE B encode hydroxyalkyl a-pyrone synthasesrequired for pollen development and sporopollenin biosynthesis inArabidopsis thaliana. Plant Cell 22: 4045–4066.

Kuromori, T., Ito, T., Sugimoto, E., and Shinozaki, K. (2011). Arabi-dopsis mutant of AtABCG26, an ABC transporter gene, is defective inpollen maturation. J. Plant Physiol. 168: 2001–2005.

Lallemand, B., Erhardt, M., Heitz, T., and Legrand, M. (2013). Sporo-pollenin biosynthetic enzymes interact and constitute a metabolon

localized to the endoplasmic reticulum of tapetum cells. Plant Physiol.162: 616–625.

McFarlane, H.E., Shin, J.J.H., Bird, D.A., and Samuels, A.L. (2010).Arabidopsis ABCG transporters, which are required for export ofdiverse cuticular lipids, dimerize in different combinations. PlantCell 22: 3066–3075.

Morant, M., Jørgensen, K., Schaller, H., Pinot, F., Møller, B.L.,Werck-Reichhart, D., and Bak, S. (2007). CYP703 is an ancientcytochrome P450 in land plants catalyzing in-chain hydroxylation oflauric acid to provide building blocks for sporopollenin synthesis inpollen. Plant Cell 19: 1473–1487.

Morgan, T.E., and Huber, G.L. (1967). Loss of lipid during fixation forelectron microscopy. J. Cell Biol. 32: 757–760.

Niu, B.-X., He, F.-R., He, M., Ren, D., Chen, L.-T., and Liu, Y.-G. (2013).The ATP-binding cassette transporter OsABCG15 is required for antherdevelopment and pollen fertility in rice. J. Integr. Plant Biol. 55: 710–720.

Palade, G.E. (1952). A study of fixation for electron microscopy. J.Exp. Med. 95: 285–298.

Piffanelli, P., Ross, J.H., and Murphy, D.J. (1998). Biogenesis andfunction of the lipidic structures of pollen grains. Sex. Plant Reprod.11: 65–80.

Pighin, J.A., Zheng, H., Balakshin, L.J., Goodman, I.P., Western,T.L., Jetter, R., Kunst, L., and Samuels, A.L. (2004). Plant cu-ticular lipid export requires an ABC transporter. Science 306: 702–704.

Prahl, A.K., Rittscher, M., and Wiermann, R. (1986). New aspects ofsporopollenin biosynthesis. In Biotechnology and Ecology of Pollen,D.L. Mulcahy, ed (New York: Springer-Verlag), pp. 313–318.

Qin, P., Tu, B., Wang, Y., Deng, L., Quilichini, T.D., Li, T., Wang, H.,Ma, B., and Li, S. (2013). ABCG15 encodes an ABC transporterprotein, and is essential for post-meiotic anther and pollen exinedevelopment in rice. Plant Cell Physiol. 54: 138–154.

Quilichini, T.D., Douglas, C.J., and Samuels, A.L. (2014a). Newviews of tapetum ultrastructure and pollen exine development inArabidopsis thaliana. Ann. Bot. (Lond.) 114: 1189–1201.

Quilichini, T.D., Friedmann, M.C., Samuels, A.L., and Douglas, C.J.(2010). ATP-binding cassette transporter G26 is required for malefertility and pollen exine formation in Arabidopsis. Plant Physiol.154: 678–690.

Quilichini, T.D., Grienenberger, E., and Douglas, C.J. (2014b). Thebiosynthesis, composition and assembly of the outer pollen wall: Atough case to crack. Phytochemistry, in press.

Rozema, J., Broekman, R.A., Blokker, P., Meijkamp, B.B., de Bakker,N., van de Staaij, J., van Beem, A., Ariese, F., and Kars, S.M. (2001).UV-B absorbance and UV-B absorbing compounds (para-coumaricacid) in pollen and sporopollenin: The perspective to track historic UV-Blevels. J. Photochem. Photobiol. B 62: 108–117.

de Azevedo Souza, C., Kim, S.S., Koch, S., Kienow, L., Schneider,K., McKim, S.M., Haughn, G.W., Kombrink, E., and Douglas, C.J.(2009). A novel fatty acyl-CoA synthetase is required for pollendevelopment and sporopollenin biosynthesis in Arabidopsis. PlantCell 21: 507–525.

Tang, L.K., Chu, H., Yip, W.K., Yeung, E.C., and Lo, C. (2009). Ananther-specific dihydroflavonol 4-reductase-like gene (DRL1) isessential for male fertility in Arabidopsis. New Phytol. 181: 576–587.

Wang, A., Xia, Q., Xie, W., Datla, R., and Selvaraj, G. (2003). Theclassical Ubisch bodies carry a sporophytically produced structuralprotein (RAFTIN) that is essential for pollen development. Proc.Natl. Acad. Sci. USA 100: 14487–14492.

Wehling, K., Niester, C., Boon, J.J., Willemse, M.T., and Wiermann,R. (1989). p-Coumaric acid—A monomer in the sporopolleninskeleton. Planta 179: 376–380.

Polyketide Trafficking in Exine Formation 4497

Weng, J.-K., Li, Y., Mo, H., and Chapple, C. (2012). Assembly of anevolutionarily new pathway for a-pyrone biosynthesis in Arabi-dopsis. Science 337: 960–964.

Weng, J.-K., Mo, H., and Chapple, C. (2010). Over-expression of F5Hin COMT-deficient Arabidopsis leads to enrichment of an unusuallignin and disruption of pollen wall formation. Plant J. 64: 898–911.

Xu, J., Yang, C., Yuan, Z., Zhang, D., Gondwe, M.Y., Ding, Z., Liang,W., Zhang, D., and Wilson, Z.A. (2010). The ABORTED MICRO-SPORES regulatory network is required for postmeiotic male

reproductive development in Arabidopsis thaliana. Plant Cell 22: 91–107.

Yamaguchi, T., Kurosaki, F., Suh, D.Y., Sankawa, U., Nishioka, M.,Akiyama, T., Shibuya, M., and Ebizuka, Y. (1999). Cross-reactionof chalcone synthase and stilbene synthase overexpressed inEscherichia coli. FEBS Lett. 460: 457–461.

Zipfel, W.R., Williams, R.M., and Webb, W.W. (2003). Nonlinearmagic: Multiphoton microscopy in the biosciences. Nat. Biotechnol.21: 1369–1377.

4498 The Plant Cell

DOI 10.1105/tpc.114.130484; originally published online November 21, 2014; 2014;26;4483-4498Plant Cell

Teagen D. Quilichini, A. Lacey Samuels and Carl J. DouglasArabidopsisPollen Wall Exine Formation in

ABCG26-Mediated Polyketide Trafficking and Hydroxycinnamoyl Spermidines Contribute to

 This information is current as of September 28, 2020

Supplemental Data /content/suppl/2014/11/24/tpc.114.130484.DC1.html

References /content/26/11/4483.full.html#ref-list-1

This article cites 54 articles, 24 of which can be accessed free at:

Permissions https://www.copyright.com/ccc/openurl.do?sid=pd_hw1532298X&issn=1532298X&WT.mc_id=pd_hw1532298X

eTOCs http://www.plantcell.org/cgi/alerts/ctmain

Sign up for eTOCs at:

CiteTrack Alerts http://www.plantcell.org/cgi/alerts/ctmain

Sign up for CiteTrack Alerts at:

Subscription Information http://www.aspb.org/publications/subscriptions.cfm

is available at:Plant Physiology and The Plant CellSubscription Information for

ADVANCING THE SCIENCE OF PLANT BIOLOGY © American Society of Plant Biologists