TURAN and EVAN Mediate Pollen Tube Reception in Arabidopsis Synergids through Protein Glycosylation

Pollen tube (PT) reception in flowering plants describes the crosstalk between the male and female gametophytes upon PT arrival at the synergid cells of the ovule. It leads to PT growth arrest, rupture, and sperm cell release, and is thus essential to ensure double fertilization. Here, we describe TURAN (TUN) and EVAN (EVN), two novel members of the PT reception pathway that is mediated by the FERONIA (FER) receptor-like kinase (RLK). Like fer, mutations in these two genes lead to PT overgrowth inside the female gametophyte (FG) without PT rupture. Mapping by next-generation sequencing, cytological analysis of reporter genes, and biochemical assays of glycoproteins in RNAi knockdown mutants revealed both genes to be involved in protein N-glycosylation in the endoplasmic reticulum (ER). TUN encodes a uridine diphosphate (UDP)-glycosyltransferase superfamily protein and EVN a dolichol kinase. In addition to their common role during PT reception in the synergids, both genes have distinct functions in the pollen: whereas EVN is essential for pollen development, TUN is required for PT growth and integrity by affecting the stability of the pollen-specific FER homologs ANXUR1 (ANX1) and ANX2. ANX1- and ANX2-YFP reporters are not expressed in tun pollen grains, but ANX1-YFP is degraded via the ER-associated degradation (ERAD) pathway, likely underlying the anx1/2-like premature PT rupture phenotype of tun mutants. Thus, as in animal sperm–egg interactions, protein glycosylation is essential for the interaction between the female and male gametophytes during PT reception to ensure fertilization and successful reproduction.


Introduction
In flowering plants, male and female gametes are constituents of the male (pollen) and female gametophytes (FG, embryo sac).A complex series of communication events between the male gametophyte and the female tissues of the flower is required for the pollen tube (PT) to deliver the two immotile sperm cells to the FG [1].During double fertilization, one sperm cell each fuses with the egg cell and central cell to give rise to the embryo and endosperm, respectively.To reach the ovule-embedded FG, the PT grows through the stigma, the style, and the transmitting tract, enters the ovary, and grows along the funiculus towards the ovule.Gradients of various small, organic molecules [2,3], as well as larger peptides produced by the synergids [4], play an essential role in directing the PT to the FG.The synergids flank the egg cell at the micropylar end of the FG and secrete LUREs, small defensin-like proteins (DEFLs), which form a subgroup of cysteine-rich polypeptides (CRPs) [4][5][6].Transduction of these female signals in the PT involves two receptor-like cytoplasmic kinases, LOST IN POLLEN TUBE GUID-ANCE1 (LIP1) and LIP2 [7].After arrival of the PT at the micropyle, it grows beyond the filiform apparatus (FA), a membrane-rich structure at the micropylar pole of the synergids, enters the receptive synergid, and ruptures to release the sperm cells [8].Therefore, interactions between the PT and the synergids might consist of two spatially and temporally distinct stages.The first being PT reception at the FA, where PT growth is temporally slowed down or arrested, and the second involving rapid growth towards the PT entry site, PT rupture, and release of the two sperm cells with the concomitant death of the receptive synergid [9].FERONIA (FER), a receptor-like kinase (RLK) of the Catharanthus roseus receptor-like kinase 1-like (CrRLK1L) subfamily [10,11], is localized to the FA [12].In mutants lacking FER activity, synergid development is normal, but fer (or sirène [13], which is allelic) mutant FGs remain unfertilized, even by wild-type PTs.The fer mutant thus revealed that an active signaling process is required for PT reception [10,12,13].In these unfertilized ovules, the PT enters the receptive synergid but neither stops its growth nor ruptures to release the sperm cells.
Instead, PT growth continues inside the FG leading to a PT overgrowth phenotype [10,12,13].The two closest homologs of FER are the pollen-specific genes ANXUR1 (ANX1) and ANX2 [14,15].anx1 and anx2 single mutants have no phenotype, but anx1/2 double mutant PTs burst in vitro and in vivo shortly after germination [14,15].Signaling via ANX proteins activates NADPH oxidases that produce reactive oxygen species (ROS) [16].They fine-tune the Ca 2+ -gradient at the PT tip, resulting in the sustained secretion of membrane and cell wall material required for steady PT elongation [16].However, upon PT arrival at the FG, FER-dependent accumulation of ROS around the FA leads to PT rupture [17].Thus, the two ANX RLKs seem to ensure PT integrity until its arrival at the FA, and activation of the FER-dependent PT reception pathway leads to ROS accumulation and PT rupture [14,15,17].
Recently, binding of the 5kDa small peptide RAPID ALKANIZATION FACTOR1 (RALF1) to FER was shown in roots, where it leads to phosphorylation of plasma-membrane H + -ATPase 2 that regulates cell elongation [18].However, it remains unclear whether pollen-expressed RALF-like proteins bind to FER in the synergids and activate the PT reception signaling cascade.In addition to FER, the following factors were shown to play a role in PT reception: LORELEI (LRE), a synergid-expressed glycosylphosphatidylinositol (GPI)-anchored protein [19], and NORTIA (NTA), a mildew-resistance locus O (MLO) family protein that accumulates at the FA in a FER-dependent manner upon PT arrival [20].lre-1/LRE and nta-1/ nta-1 mutants show the fer-like PT overgrowth phenotype in 28% and 22% of the ovules, respectively [19,20].Furthermore, the abstinence by mutual consent (amc) mutant, which affects a peroxin involved in protein import into peroxisomes, shows the fer-like phenotype only if mutant PTs contact mutant FGs [21].This particular phenotype suggests a disrupted communication between both gametophytes due to missing signaling molecules from peroxisomes, possibly ROS.Recently, the first male gametophytic factors affecting PT reception were identified [22,23].PTs of a triple mutant disrupting three MYB transcription factors (myb97-1, myb101-4, and myb120-3) fail to rupture and release the sperm in 60-70% of targeted ovules [22,23].However, the target genes of these transcription factors, which may be involved in signaling, remain to be identified.
Here, we describe two novel mutants impaired in PT reception, turan (tun) and evan (evn).Interestingly, tun/TUN and evn/EVN plants show the same female gametophytic fer-like PT overgrowth phenotype but have distinct pollen defects.Whereas evn mutant pollen grains degenerate before maturation, tun mutant grains develop normally, but PTs burst immediately after in vitro germination, reminiscent of the anx1/2 phenotype.TUN and EVN are both involved in protein N-glycosylation in the endoplasmic reticulum (ER) and encode a putative uridine diphosphate (UDP)-glycosyltransferase superfamily protein and a dolichol kinase, respectively.The mutants do not affect the abundance and subcellular localization of FER, NTA, and LRE fusion proteins, suggesting that aberrant glycosylation may affect the function of at least one of these membrane proteins.In tun mutant pollen grains, ANX1 fused to the yellow fluorescent protein (ANX1-YFP) is degraded by the ER-associated degradation (ERAD) pathway.This leads to premature PT rupture, indicating that the ANX1/2 RLKs are targets of TUN-dependent N-glycosylation.

tun and evn Mutants Show a fer-like Pollen Tube Overgrowth Phenotype
To gain more insight into the molecular mechanisms involved in PT reception in Arabidopsis, we conducted a forward genetic screen that yielded several mutants showing a fer-like PT overgrowth phenotype.The three mutants with the highest penetrance were chosen for further characterization, but two turned out to be allelic to each other (see below).Like fer, both mutants were named after Etruscan goddesses of fertility and fate, namely turan (tun) and evan (evn) [24].In tun-1/TUN, evn-1/EVN, and evn-2/EVN heterozygous mutants 12% (n = 1,318), 20% (n = 1,233), and 22% (n = 320) of the ovules remained unfertilized, respectively, compared to only 1.5% (n = 1,389) in wild-type plants (Table 1).In unfertilized ovules, the PT continued to grow inside the FG, failed to arrest its growth, and did not rupture to release the sperm cells (Fig . To ensure that the PT overgrowth phenotype was not caused by cell specification defects in these mutants, a β-glucuronidase (GUS) synergid fate marker (ET2634) was analyzed.In both tun and evn mutant ovules, GUS expression was restricted to the synergids (Fig 1D -1F), indicating that their identity was not affected.However, some ovules showed an abnormal structure at the micropylar pole of the synergids (Fig 1E).To further investigate this structure, ovule membrane staining, clearings, and sections were analyzed 2 d after emasculation (DAE).Although membrane staining revealed no change in overall synergid morphology in the mutants (S2 Fig and S1 Data), approximately 50% of the mature FGs in cleared ovules of tun-1/TUN and evn-1/EVN mutants showed the abnormal structure in the FA region (Fig 1G -1I).Using Aniline Blue staining on ovule sections, we found that approximately 50% of the mature FGs showed increased callose deposition at the micropylar pole in both mutants (Fig 1J -1L).However, this did not influence PT attraction and reception, since all ovules could attract PTs, and over 60% of mutant FGs were fertilized.To investigate whether callose deposition in tun and evn FGs is an indicator of a defense-related response [25], expression of several plant defense pathway genes was tested in mutant pistils 2 DAE, but no up-regulation was observed (S3 Fig).
In summary, we identified two novel members of the PT reception pathway in the synergids, which are required for successful reproduction.In both mutants, synergid differentiation is normal, but callose accumulates at the micropylar pole of mutant FGs that, however, does not mediate the failure in PT reception.

tun and evn Mutants Show Additional but Distinct Male Gametophytic Defects
Self-pollination of both tun-1/TUN and evn-1/EVN mutant plants yielded only heterozygous and wild-type offspring.Attempts to propagate the mutants by crossing wild-type plants with mutant pollen produced no mutant progeny either (n = 96 plants per mutant; Table 1).These results suggest that not only the female but also the male gametophyte is affected in the tun and evn mutants.To further investigate this hypothesis, tun-1/TUN and evn-1/EVN plants were crossed to quartet (qrt/qrt) mutants [26,27], where microspores fail to separate after meiosis, forming tetrads of pollen grains.Heterozygous mutants in a qrt/qrt background facilitate the analysis of pollen defects because, within one tetrad, two microspores carry the wildtype and two the mutant allele.In vitro pollen germination experiments revealed that in tun-1/ TUN mutants, pollen development was normal, but 63% ± 1.5% (n = 441) of mature PTs burst immediately after germination, compared to 5% ± 5% (n = 128) in the wild type (Fig 2A -2C and Table 1).This pollen phenotype of tun is reminiscent of double mutants disrupting ANX1/ 2, the pollen-specific homologs of FER [14,15].In contrast to the wild type, where nearly all pollen were intact at maturity (n = 600), 54% ± 5% (n = 800) of the pollen grains degenerated before maturation in evn-1/EVN mutants, indicating general defects in pollen development (Fig 2D and 2E and Table 1).Reciprocal crosses between tun-1/TUN and evn-1/EVN mutants with Col-0 wild-type plants showed that the fer-like PT overgrowth phenotype was caused by a female gametophytic defect.Whereas mutant ovules fertilized by wild-type pollen showed the fer-like PT overgrowth phenotype, wild-type pistils pollinated with pollen from heterozygous mutants showed no phenotype (S5 Fig) as expected, since no functional mutant PTs are formed.Both evn and tun pollen defects were fully penetrant (i.e., affecting male gametophyte viability and PT growth, respectively) with a male transmission efficiency of 0% (n = 96 plants per mutant; Table 1).In contrast, female transmission efficiency was reduced to 74.5% (n = 96) and 28% (n = 96) in tun-1/ TUN and evn-1/EVN mutants, respectively (Table 1).Taken together, these results show that tun and evn mutants display distinct male gametophytic defects, affecting male gametophyte growth and viability, respectively.However, abnormal PT reception is caused by a female gametophytic defect, leading to the decreased female transmission of the mutant alleles.

TUN Encodes a Putative UDP-Glycosyltransferase and EVN a Dolichol Kinase
In order to identify the causative mutations, we developed SNP-ratio mapping (SRM) [28].SRM enables the mapping of heterozygous mutant individuals by next-generation sequencing.Briefly, the segregation ratios of EMS-induced SNPs are used to identify the causative SNP that segregates in a 1:1 ratio in a backcross mutant population versus a 1:3 ratio for unlinked SNPs segregating in the background.
Applying SRM to tun-1/TUN mutants revealed a stop codon in the sixth exon of At1g16570 (Fig 2F) [28], which encodes a putative UDP-glycosyltransferase superfamily protein belonging to the glycosyltransferase (GT) family 33, whereof TUN is the only member in Arabidopsis.In evn-1/EVN mutants, SRM identified a stop codon in the eleventh exon of At3g45040 (  ), which encodes the only dolichol kinase in the Arabidopsis genome [29].Thus, both TUN and EVN encode proteins that are likely playing roles in protein N-glycosylation.Whereas EVN as the only dolichol kinase may have a general role, TUN potentially acts in a more specific manner, since a total of 27 GT families have been identified in Arabidopsis.Finally, a second allele of evn (evn-2) was found in the same EMS screen and identified by a combination of SRM (S7  ).
To confirm that the correct genes had been identified, T-DNA insertion lines disrupting them were analyzed.As described previously, the T-DNA allele tun-2 (SAIL_400_A01) has an insertion in the fourth exon of At1g16570 and displays a PT overgrowth phenotype (15%; n = 513; Fig 2F and Table 1 and S1B Fig) [28] and PT bursting in vitro (44% ± 6.5%; n = 404; Table 1 and S1F  In summary, TUN and EVN both encode proteins potentially involved in protein N-glycosylation, suggesting that proper N-glycosylation of proteins involved in PT reception is critical for the gametophytic dialogue during PT reception.

TUN and EVN Are Involved in Protein N-Glycosylation, and Down-Regulation Causes Distinct Vegetative Phenotypes
N-linked protein glycosylation occurs as proteins transit through the ER [30].To gain insight into the subcellular localization of the TUN and EVN proteins, and to confirm their identity by functional complementation, native promoter::protein-green fluorescent protein (GFP) fusions were transformed into mutant plants.The pTUN::TUN-GFP construct complemented both the female and the male phenotypes (Table 1).In ovules, the strongest GFP signal was observed in the FG including the synergids, which showed a ring shaped signal around their nuclei, and throughout the PT (S8A-S8C Fig) .The pEVN::EVN-GFP construct neither showed GFP expression nor complemented the phenotypes, suggesting that the construct was not functional in planta, likely due to missing regulatory elements.As no homozygous mutants could be recovered, we used RNA interference (RNAi) to down-regulate expression of TUN and EVN using the 35S promoter.Four independent TUN(RNAi) and EVN(RNAi) lines with greatly reduced expression, down to 12% and 13% of wild-type levels, respectively (S9 Fig and S1 Data), were chosen for further analysis.To investigate the function of TUN and EVN in a biochemical analysis of glycoproteins, potential changes in glycoprotein abundance in knockdown seedlings were assessed by a lectin blot using Concanavalin A (ConA), which mainly binds to terminal mannosyl and glucosyl residues of glycoproteins.Both, TUN(RNAi) and EVN(RNAi) seedlings showed altered glycoprotein abundance and mobility compared to both the wild type and ost3/6-2, a mutant disrupted in a subunit of an oligosaccharyltransferase that acts later in the glycosylation pathway (Fig 3) [31].
Furthermore, TUN(RNAi) knockdown lines displayed a vegetative phenotype reminiscent of fer/fer mutants (Fig 4) [20,32]  Taken together, both TUN and EVN play a role in protein N-glycosylation but only TUN seems required for normal vegetative growth and development.

Down-Regulation of TUN Does Not Cause Deglycosylation of FER
Protein N-glycosylation plays important roles in many processes, including protein folding, protein stabilization, protein targeting [33], and receptor-ligand interactions [34].In tun mutants, we observed fer-like PT overgrowth, a fer-like vegetative phenotype, and anx1/2-like PT bursting.The FER, ANX1, and ANX2 proteins have multiple predicted glycosylation sites [11], indicating that TUN may be involved in the specific glycosylation of this RLK subfamily.To investigate whether FER requires TUN for glycosylation, a TUN(RNAi) construct was transformed into plants expressing a pFER::FER-GFP translational fusion protein.In the case of a complete absence of potential N-glycans attached to FER, we would expect a smaller size of the under-glycosylated FER protein in TUN(RNAi) seedlings compared to the wild type [31].However, no obvious difference in size, indicating a loss of glycosylation, was detected between FER-GFP in TUN(RNAi) and wild-type seedlings by immunoblot analysis using an anti-GFP antibody (S10 Fig) .In contrast, a clear shift in size was seen when FER-GFP was deglycosylated by Endoglycosidase H (EndoH, a high mannose N-glycan deglycosylase) treatment in vitro (S10 Fig).
These data suggest that FER-GFP is not completely deglycosylated in TUN(RNAi) seedlings.However, we cannot exclude the possibility that residual levels of TUN activity in the RNAi lines were able to partially glycosylate FER in seedlings, or that FER-GFP is misglycosylated rather than deglycosylated.Although no reporter gene was available for LRE, encoding a predicted GPI-anchored protein, in situ hybridization showed LRE expression predominantly in synergids before fertilization [19].A LRE-GFP fusion protein did not produce a detectable fluorescent signal in transient transformations [19], possibly due to the exposure of GFP to the acidic pH in the apoplast.However, LRE is a good candidate for a target of EVN because the yeast mutant corresponding to evn, secretory59 (sec59), is depleted in GPI-anchored proteins [35].Therefore, we produced a LRE reporter using the pH-stable fluorescent protein Citrine [36], and cloned it between the predicted signal peptide and GPI-anchor of LRE [19].The LRE-Citrine fusion protein localized to the surface of the synergids, appearing different from the FA localization of FER-GFP (Fig 5D and 5G).Thus, LRE-Citrine likely faces the extracellular space towards the micropyle, where PT reception is initiated upon PT arrival.However, LRE-Citrine production and localization were unaffected in evn and tun mutants ( In summary, PT reception defects in tun and evn FGs are not caused by misexpression or mislocalization of FER, NTA, and LRE.However, it is possible that protein function is impaired due to misglycosylated residues-particularly in FER.
Thus, ANX1-YFP and ANX2-YFP are not detectable in tun mutant pollen grains, possibly because they are subjected to ERAD of misfolded proteins [37].Indeed, treatment with Kifunensine (Kif), an inhibitor of the ERAD pathway [38] resulted in the recovery of ANX1-YFP fluorescence in tun pollen grains (Fig 6D).Because the outcomes of Kif treatments were variable, likely dependent on how well the drug was taken up in different experiments, we also used the ERAD inhibitor Eeyarestatin I (EerI) [39], which resulted in a more consistent recovery of ANX1-YFP fluorescence ( Therefore, TUN seems to be involved in ANX1-YFP and ANX2-YFP glycosylation, and the aberrant glycosylation of these fusion proteins leads to their degradation by the ERAD pathway.However, although less likely, we cannot exclude that ANX1/2-YFP proteins get degraded because another protein required for their stability is misglycosylated and subject to the ERAD pathway.

TUN and EVN Encode Proteins of the Early N-Glycosylation Pathway
In this study, we characterized two mutants, tun and evn, isolated in a screen for defects in PT reception.They have similar PT overgrowth phenotypes in the FG, but play distinct roles during pollen development, PT integrity, and vegetative growth.Both genes encode proteins that are involved in N-linked protein glycosylation, indicating that this cotranslational protein modification is essential to various developmental processes.
N-linked glycosylation affects protein folding, stability, transport, and activity [40].It is a multistep process starting with the assembly of an oligosaccharide, containing N-acetylglucosamine (GlcNAc), mannose (Man), and glucose (Glc), on a phosphorylated membrane-bound polyisoprenoid lipid carrier, dolichol [30].The assembly starts on the cytosolic side and finishes on the lumenal side of the ER, yielding the end product Glc 3 -Man 9 -GlcNAc 2 [30].The second step is the cotranslational transfer of the oligosaccharide to an asparagine in the Asn-X-Ser/Thr sequon, where X can be any amino acid except proline [41].The last step is ER-associated quality control, which ensures proper folding of N-glycosylated proteins and their subsequent exit of the ER [37].
EVN encodes a dolichol kinase, which plays an early role in the N-glycosylation pathway [29].The SEC59 dolichol kinase in yeast catalyzes the cytidine-triphosphate-dependent phosphorylation of ER-membrane-bound dolichol [35].As EVN can complement the yeast sec59 mutant, it seems to perform the same biochemical function [29].Phosphorylated dolichol (Dol-P) serves as carrier for the assembly of the oligosaccharide on the cytosolic side of the ER and, additionally, as a carrier for the monosaccharides Man and Glc in the ER lumen for GPIanchor synthesis.In Arabidopsis, EVN is a single copy gene with characterized orthologs in yeast [42], mammals [43,44], and humans [45], where dolichol kinase deficiency results in hypoglycosylation and lethality [46,47].We never retrieved any homozygous mutants for evn, due to its complete male gametophytic lethality.As EVN(RNAi) knockdown plants do not show any obvious vegetative phenotype, a low, residual gene activity might be sufficient for normal plant growth.
TUN encodes a putative UDP-glycosyltransferase superfamily protein belonging to GT family 33, whereof TUN is the only member in Arabidopsis.The yeast ortholog ALG1 encodes a beta-1,4-mannosyltransferase, which transfers the first Man from the GDP-D-Man substrate to the Dol-PP-GlcNAC 2 acceptor on the cytosolic side of the ER [48,49].ALG1 resides in the ER membrane, forming homodimers and heterodimers with ALG2 and ALG11, respectively [50].Under nonpermissive conditions, temperature-sensitive alg1-1 mutants produce no Mancontaining oligosaccharides [48], leading to lethality [51].As for evn, no homozygous tun mutants were recovered because of complete male gametophytic lethality.However, the strong dwarf phenotype and dying seedlings in TUN(RNAi) lines indicate that TUN is also required for vegetative growth and development.In plants, the importance of protein N-glycosylation during embryogenesis, vegetative development, and plant defense has been demonstrated by various phenotypes of N-glycosylation mutants [52].
The subcellular localization of TUN and EVN fusion proteins in the ER and the altered glycoprotein patterns observed in RNAi lines for both genes indicate that the plant proteins play similar roles in protein glycosylation as their yeast orthologs.

Localization of Known Members of the Pollen Tube Reception Pathway Is Not Affected in evn and tun Synergids
We showed that mutations in EVN and TUN in the FG lead to PT reception defects.Due to their protein localization to the ER, it is likely that EVN and TUN do not directly mediate male-female gametophyte interactions during PT reception, but that an impaired glycosylation pathway affects the function of female players involved in this process.However, the localization of translational reporter gene fusions with FER, NTA, and LRE, three known components of the PT reception pathway, was unaffected in FGs of evn-3/EVN and tun-2/TUN plants.This was not surprising for NTA-GFP, which has no predicted glycosylation sites.However, LRE-Citrine fluorescence was expected to be reduced or even absent in evn mutant ovules, since the yeast sec59 mutant is depleted in GPI-anchored proteins [35].However, EVN and SEC59 display only 22% amino acid identity (ClustalW).Although the plant protein is able to complement the yeast sec59 mutant [29], the two proteins may not play identical roles in planta and/ or other, undescribed kinases could take over EVN function in mutant plants.
However, normal protein abundance and localization does not ensure proper function.The fact that tun mutants show fer-like PT overgrowth, a fer-like vegetative phenotype, and anx1/2like PT bursting, suggests that the FER and ANX1/2 RLKs are affected in this mutant.FER has nine putative N-glycosylation sites, eight of which reside in the extracellular malectin-like domain [11].In Xenopus laevis, malectin is an ER-localized lectin, which selectively binds carbohydrates and is involved in the ER-quality control of glycoproteins [53,54].The extracellular, glycosylated malectin-like domain of FER suggests that its ligand could contain sugar residues and/or that proper glycosylation of this domain is essential for ligand binding.However, potential deficiencies in protein glycosylation in tun and evn mutant FGs did not affect polar localization of FER-GFP to the FA, and down-regulation of TUN did not cause a complete loss of Nlinked glycans in FER-GFP in seedlings.But it is possible that only some N-glycans are absent or that N-glycan composition is altered.Therefore, while FER-GFP localization is unaffected, functions such as ligand binding and/or recognition and subsequent signal transduction could be affected.This has been shown to be the case for the plant immunity leucine-rich repeat (LRR) RLK EF-Tu receptor (EFR), which shows impaired ligand binding as a consequence of underglycosylation [34].Just recently, binding of the 5kDa small peptide RALF1 to FER was shown in roots [18].However, RALF1 is neither expressed in PTs nor in synergid cells.But there are 34 RALF-like peptides with various expression patterns throughout the plant [18].In the case of RALF1, N-glycosylation does not seem to play a role in FER binding, since the processed peptide lacks any predicted N-glycosylation site.Out of seven highly pollen-expressed RALF-like peptides, RALF-like4 and RALF-like26 are the only two with a predicted N-glycosylation site.It is conceivable that, unlike in animals [53,54], the malectin-like ligand-binding domain of FER has no conserved carbohydrate binding capacity in plants.Instead, the glycosylation state of the ligand-binding domain of the receptor could be more important than glycosylation of the ligand.
PT reception at the FA is essential for double fertilization and thus successful reproduction.It is conceivable that a "dual recognition system" exists, where the protein backbone of the receptor is required for interactions with the ligand, which are further enhanced by N-linked glycans, thereby increasing the chances for successful PT reception.Such a dual recognition system has been described for gamete interactions in mammals [55], where the glycoprotein mZP3 in the egg cell's extracellular matrix, the zona pellucida, is responsible for sperm binding [56].While sperm binding is improved by glycosylated mZP3 proteins, unglycosylated mZP3 can also bind sperm.Accordingly, a sperm protein or protein complex interacts with the glycans and/or the protein backbone of mZP3 depending on its glycosylation state.Thus, this dual adhesion system ensures better sperm-egg binding but allows gamete interaction even if glycan composition is disturbed, increasing the chance for successful fertilization [57].In plants, the existence of such a dual recognition system could explain the reduced penetrance of the PT overgrowth phenotype in tun and evn since the female protein component, even if not glycosylated properly, could still partially recognize the male ligand.Despite the similarities of the fer and tun phenotypes, further biochemical investigations are required to investigate whether indeed the FER RLK and/or other components of the PT reception pathway are affected in the tun mutant.

ANX1-YFP Gets Degraded by the ERAD Pathway in tun Pollen Grains
Apart from the phenotypes it shares with fer, tun also shows anx1/2-like PT bursting.ANX1 and ANX2 have seven and four potential N-glycosylation sites, respectively [11], with all four predicted glycosylation sites of ANX2 being located in the extracellular domain and conserved in ANX1.ANX1-YFP and ANX2-YFP fluorescence is not detectable in tun-2 mutant pollen grains, but ANX1-YFP fluorescence is recovered after treatment with two ERAD inhibitors.This finding suggests that ANX1-YFP is misglycosylated in tun pollen grains and is therefore degraded by the ERAD pathway, which is induced when N-glycan-dependent protein folding fails [37].The misfolded protein is recognized by ubiquitin ligases, ubiquitinatinylated, retrotranslocated to the cytosol, and degraded by the proteasome [37].For instance, the LRR-RLK BRASSINOSTEROID-INSENSITIVE1, which contains multiple N-glycosylation sites [58], is such an N-glycan-dependent ERAD target [59,60].
Since FER, NTA, and LRE fusion proteins did not show any alteration in protein abundance or localization in tun mutant ovules, the effect on ANX1-YFP and ANX2-YFP seems to be a pollen-specific effect rather than a general loss of glycosylated proteins in the plant.Specificity of ERAD has been described for the LRR-RLKs EFR and FLAGELLIN-SENSITIVE2 (FLS2), involved in plant innate immunity [61].Both LRR-RLKs are N-glycosylated membrane proteins, but mutations in two members of the ER quality control pathway affected only EFR but not FLS2 [61].Similarly, the closely related CrRLK1Ls FER, ANX1, and ANX2 might all be misglycosylated in the tun mutant, but only ANX1/2 are degraded via the ERAD pathway, explaining the absence of ANX1/2-YFP but the normal abundance of FER-GFP.This would also explain the difference in the penetrance of male and female gametophytic phenotypes; although FER is likely misglycosylated, it is still localized to the FA, thereby allowing successful PT reception to a certain degree, consistent with the residual transmission of tun through the FG.In contrast, ANX1/2 are degraded by the ERAD pathway leading to a fully penetrant anx1/ 2-like phenotype and the complete absence of transmission through the pollen.

Conclusions
The characterization of TUN and EVN demonstrated that protein N-glycosylation is important for various developmental processes.Impaired PT reception and anx1/2-like PT rupture in tun mutants appears to be linked to misglycosylated FER and ANX1/2, respectively, the most closely related members of the CrRLK1L subfamily of RLKs.The defects in PT reception observed in the tun and evn glycosylation mutants provides a first indication that plants evolved similar mechanisms to ensure fertilization as mammals, where both N-glycan-protein and proteinprotein interactions appear to act synergistically to guarantee gamete binding and, thus, enhance the chance for successful fertilization.

Aniline Blue Staining
To visualize PTs, siliques were selected around 2 DAP.Sepals and petals were removed, and siliques were fixed in 9:1 ethanol (EtOH):acetic acid overnight at 4°C.Aniline Blue staining was previously described [10], and sample analysis was done using a Leica DM6000B epifluorescence microscope.Callose staining of semithin ovule sections is described in S1 Text.

In Vitro Pollen Germination Analysis
In vitro pollen germination was previously described [14].Pollen grains and tubes were imaged using the differential interfering contrast on a Leica DM6000B microscope.

FM4-64 Membrane Staining
Wild-type, tun-2/TUN, and evn-3/EVN pistils were dissected in ice cold 100 mM FM4-64 Dye (Life Technologies) solution on microscope slides, and ovules were covered with a cover slip.Microscope slides were covered with aluminum foil and left on ice for 3-4 h.Fluorescence of ovule membranes was analyzed using a Leica SP5 confocal microscope.

Confocal Microscopy
Confocal microscopy was previously described [20], with the exception that a Leica SP5 confocal microscope was used.

Kifunensine Treatment
Whole inflorescences were cut off the plant and incubated in 50 μM Kif solution (Sigma-Aldrich).After 2 d under constant light at 22°C, fluorescence in mature pollen grains was analyzed using a Leica SP5 confocal microscope.

Eeyarestatin I Treatment
Anthers of flower buds (around stage 11) were dissected and pollen grains were placed on pollen germination media [14] containing various concentrations of EerI (Sigma).The pollen was incubated for 20 h at 22°C in a moisture incubation box.Fluorescence in pollen grains was analyzed using a Leica SP5 confocal microscope.The SNP region of the At3g45040 gene was amplified from each of the 94 and 60 DNA samples that had been pooled for sequencing from evn-1 and evn-2, respectively, and two Col-0 controls.PCR products were digested with the SURVEYOR nuclease, cleaving single base pair mismatches in heteroduplex DNA [64].In evn-1, the undigested wild-type band is 1,000 bp, whereas any sample containing the SNP displays an undigested band at 1,000 bp and two digestion products at around 800 bp and 200 bp.Individuals six and seven have a wild-type band only, and were shown to be sampling mistakes.In evn-2, the undigested wild-type band is 900 bp, whereas any sample containing the SNP displays an undigested band at 900 bp and two digestion products at around 500 bp and 400 bp.Results for tun-1 were published previously [28].(PDF)

Supporting Information
S1 Table .SNP data of the resequenced, EMS-mutagenized genome of evn-1/EVN.All data presented has been obtained using the DiBayes algorithm for SNP calling with the highest stringency settings [28].(XLS) S2 Table .SNP data of the resequenced, EMS-mutagenized genome of evn-2/EVN.All data presented has been obtained using the DiBayes algorithm for SNP calling with the highest stringency settings [28].(XLS) S3 Table.Primers to amplify evn2 linked genes carrying a non-causative EMS SNP for SUR-VEYOR analysis.SNP containing regions of 14 genes were amplified.The segregation ratio of the SNP by SRM is indicated in brackets.

Fig 2 .
Fig 2. Distinct pollen defects in tun and evn mutants.(A) Pollen in vitro germination assay of qrt/qrt pollen grains.(B) Pollen in vitro germination assay of tun-1/TUN;qrt/qrt pollen.Arrows indicate PT bursting.(C) Graph of PT bursting counts in qrt/qrt and tun-1/TUN;qrt/qrt pollen.(D) Pollen in vitro germination assay of evn-1/EVN;qrt/qrt mutant pollen.Arrowheads indicate degenerated pollen grains.(E) Graph of degenerating pollen grain counts at different stages of evn-1/EVN mutants after DAPI staining.Stage four refers to the bicellular and early tricellular, stage three to the tricellular, stage two to the late tricellular and early mature, and stage one to the mature pollen stage.Scale bars: 20 μm.(F-G) Gene model of TUN (F) and EVN (G) with mutant alleles.Ethane methyl sulfonate (EMS) single nucleotide polymorphisms (SNPs) are indicated by lines, T-DNA insertions by triangles.doi:10.1371/journal.pbio.1002139.g002 Fig 2F and S6 Fig and S1 Table Fig) and classical map-based cloning (S1 Text).This allele has a premature stop codon in the last exon and displays similar PT overgrowth (22%; n = 320) and pollen phenotypes (50% ± 0%; n = 200) as evn-1 (Fig 2F and Table 1 and S1C Fig, S1G Fig and S7 Fig and S2 Table Fig) similar to the EMS allele tun-1, confirming the correct identification of TUN by SRM.Likewise, evn-3 (SAIL_529_E06) has an insertion in the eighth exon of At3g45040 and shows the same female (28%; n = 337) and male (50% ± 0%; n = 495) [29] phenotypes as the EMS mutants evn-1 and evn-2 (Fig 2G and Table 1 and S1D Fig and S1H Fig).
Colocalization studies of p35S::TUN-GFP and p35S::EVN-GFP, where the fusion proteins were expressed under the constitutive, viral 35S promoter, with different subcellular markers were performed in transiently transformed onion and tobacco epidermis cells.They showed ER localization of both proteins (S8D and S8E Fig), supporting potential roles of TUN and EVN in protein N-glycosylation.
. The severity and frequency of the dwarf phenotype correlated with the expression level of TUN in the TUN(RNAi) lines (Fig 4C and 4G and S9 Fig).Seedlings on plates looked normal, but in soil, some plants remained small, accumulated anthocyanins, and eventually died without further growth (Fig 4C).It is possible that TUN expression in these seedlings was very low, but it could not be determined due to early lethality.These results indicate that TUN is essential for vegetative development.In contrast, EVN(RNAi) knockdown plants showed no obvious phenotype (Fig 4D and 4H and S9 Fig), suggesting that low levels of EVN are sufficient to support normal vegetative growth.
Fig 6F-6H and S12 Fig and S1 Data).These results indicate that the premature bursting of tun PTs is caused by the absence of ANX1 and ANX2.

Table 1 .
Overview of phenotypes in tun and evn mutant plants.