EXO70A2 Is Critical for Exocyst Complex Function in Pollen Development1[OPEN]

Pollen development, pollen grain germination, and pollen tube elongation are crucial biological processes in angiosperm plants that need precise regulation to deliver sperm cells to ovules for fertilization. Highly polarized secretion at a growing pollen tube tip requires the exocyst tethering complex responsible for specific targeting of secretory vesicles to the plasma membrane. Here, we demonstrate that Arabidopsis (Arabidopsis thaliana) EXO70A2 (At5g52340) is the main exocyst EXO70 isoform in the male gametophyte, governing the conventional secretory function of the exocyst, analogous to EXO70A1 (At5g03540) in the sporophyte. Our analysis of a CRISPR-generated exo70a2 mutant revealed that EXO70A2 is essential for efficient pollen maturation, pollen grain germination, and pollen tube growth. GFP:EXO70A2 was localized to the nucleus and cytoplasm in developing pollen grains and later to the apical domain in growing pollen tube tips characterized by intensive exocytosis. Moreover, EXO70A2 could substitute for EXO70A1 function in the sporophyte, but not vice versa, indicating partial functional redundancy of these two closely related isoforms and higher specificity of EXO70A2 for pollen development-related processes. Phylogenetic analysis revealed that the ancient duplication of EXO70A, one of which is always highly expressed in pollen, occurred independently in monocots and dicots. In summary, EXO70A2 is a crucial component of the exocyst complex in Arabidopsis pollen that is required for efficient plant sexual reproduction.

The pollen tube in angiosperms represents an extremely elongated cellular structure that emerges from a rehydrated pollen grain when it lands on a stigma and delivers two sperm cells to fertilize an ovule. To reach the ovule, the pollen tube navigates through pistil tissues by highly polarized tip growth, restricted to the pollen tube apex. In this tightly regulated process, cell polarity maintenance mechanisms, addition of new membrane material, and secretion of cell wall components are essential. Intensive exocytosis is localized to the very tip, followed by a subapical domain where endocytosis takes place to recycle the excess of membranes delivered in secretory vesicles (for review, see Hepler and Winship, 2015). Other key factors essential for the pollen tube tip growth include small GTPases and a Ca 21 -signaling network that regulates actin cytoskeleton dynamics (for review, see Cai et al., 2015). The cell wall of pollen tubes has a unique structure consisting of two layers: a pectinaceous and cellulose layer secreted at the apex, and an additional callose layer deposited in regions more distant from the tip. The spatial distribution of the cell wall components plays a critical role in pollen tube morphogenesis (Chebli et al., 2012). Because pollen tubes are a simplified cell-autonomous system and exhibit highly polarized growth, they provide an excellent model in cell biology for studies of cell polarity and regulation of secretion (e.g. Chebli et al., 2013;Qin and Dong, 2015). Precisely regulated membrane trafficking is crucial not only for pollen tube growth, but also for previous stages of pollen development, namely pollen grain maturation and germination (Kang et al., 2003;Peng et al., 2011;Paul et al., 2016).
While in yeast and animals the exocyst subunits are typically encoded by a single gene, in land plants they are usually duplicated or even multiplied (Elias et al., 2003;Cvrčková et al., 2012), with extreme gene proliferation in the case of the EXO70 subunit. For example, the genome of Physcomitrella patens encodes 13 EXO70 paralogs, that of Oryza sativa 47 paralogs, and that of Arabidopsis 23 paralogs, which can be divided into three clades of ancestral land plant origin, termed EXO70.1, EXO70.2, and EXO70.3 (Cvrčková et al., 2012;Rawat et al., 2017;Žárský et al., 2020). This plantspecific multiplicity of EXO70 suggests not only functional specialization in different plant tissues or cell types, but also a presence of several variants of the exocyst complex in the same cell (Žárský et al., 2013). Indeed, some EXO70.2 and EXO70.3 clade members contribute to conventional secretion in specific cell types, whereas several other isoforms acquire functions outside the conventional exocytosis pathway, acting in exocyst subcomplexes or even independently of the exocyst complex in some cases (Kulich et al., , 2015Zhang et al., 2015;Hong et al., 2016;Pečenková et al., 2017;Synek et al., 2017). In contrast, the Arabidopsis EXO70.1 isoform EXO70A1, highly expressed in most sporophytic tissues, is the main housekeeping EXO70 subunit participating in exocytosis (Synek et al., 2006;Fendrych et al., 2010;Drdová et al., 2013). However, since EXO70A1 is not expressed in pollen, the main EXO70.1 paralog functioning in pollen has remained experimentally uncharacterized until now.
The importance of the exocyst for pollen grain germination and pollen tube growth is well documented. Homozygous Arabidopsis mutants in SEC5a/b, SEC6, SEC8, and SEC15a typically produce short and wide pollen tubes with compromised polarity, resulting in a male-specific transmission defect (Cole et al., 2005;Hála et al., 2008), and sec3a mutants cannot produce pollen tubes at all (Bloch et al., 2016). Transcriptomic and proteomic analyses documented high transcription of the EXO70A2, EXO70C1, EXO70C2, EXO70F1, EXO70H3, EXO70H5, and EXO70H6 paralogs in Arabidopsis pollen, whereby EXO70C2, EXO70C1, and EXO70A2 were the most abundant in the Arabidopsis pollen proteome (Grobei et al., 2009. The two closely related EXO70C1 and EXO70C2 paralogs from the EXO70.2 clade participate in the control of optimal pollen tube tip growth . Pollen tubes lacking EXO70C2 exhibited aberrant morphology, frequently bursting and displaying a distinct stop-and-go mode of elongation, resulting in a seriously reduced fertilization capacity. These facts point to a divergent function of EXO70C2 (and its close paralog EXO70C1) independent of the conventional exocyst function . On the other hand, the clade EXO70.1 member EXO70A2 is the evolutionarily closest paralog to EXO70A1, the main sporophytic housekeeping isoform (Cvrčková et al., 2012), which represents the best candidate for the main conventional EXO70 isoform in pollen, functioning as a part of the exocyst complex in the regulation of polarized exocytosis.
In this study, we document that disruption of EXO70A2 in Arabidopsis affects pollen maturation, significantly reduces pollen germination efficiency, and causes a severe pollen-specific transmission defect of the mutant allele, which is at least in part due to impaired pollen tube growth. GFP:EXO70A2 localizes to the nucleus and cytoplasm in developing pollen grains and later to the apical PM domain in growing pollen tubes characterized by intensive exocytosis. Interestingly, ectopic expression of EXO70A2 could substitute for the EXO70A1 function in the sporophyte, but not vice versa, indicating partial functional redundancy of these two closely related isoforms. Phylogenetic and expression analysis of EXO70 genes in the EXO70.1 clade revealed independent duplications of EXO70A in monocots and dicots and showed that the dicot EXO70A2 clade contains predominantly pollen-expressed isoforms. We conclude that Arabidopsis EXO70A2 is the main EXO70 isoform functioning as a subunit of the exocyst complex in conventional secretion in Arabidopsis pollen.

EXO70A2 Is a Member of a Dicot Pollen-Expressed EXO70.1 Clade
To clarify the evolutionary relationships among EXO70A paralogs, which comprise the sole family of the ancestral EXO70.1 clade (Synek et al., 2006;Cvrčková et al., 2012;Rawat et al., 2017), we performed a detailed phylogenetic analysis of available predicted protein sequences of members of this clade from 19 plant species covering a broad range of land plant diversity including liverworts, mosses, lycophytes, gymnosperms, the basal angiosperm Amborella trichopoda, grasses as representatives of monocots, and representatives of both asterid and rosid eudicot lineages (sequences are listed in Supplemental File S1). The sole EXO70 of the charophyte Klebsormidium flaccidum was also included as an outgroup. The resulting phylogenetic tree ( Fig. 1) clearly indicates that the EXO70A duplication took place independently in basal eudicots and in monocots (or at least in grasses), resulting in both eudicots and grasses possessing representatives of two EXO70A clades that cannot be considered orthologous between the two angiosperm lineages. The Arabidopsis EXO70A1 and EXO70A2 paralogs map into different eudicot EXO70A clades, whereas the third paralog, EXO70A3, is closely related to EXO70A2 and resulted from a recent gene duplication that may have been restricted to Brassicales, or even to some subgroup thereof.
For selected representatives of both monocots and eudicots, transcript level data for the analyzed genes were extracted from public databases and mapped onto the phylogenetic tree (Fig. 1). The results document cladespecific distinct expression patterns in both eudicots and grasses. In eudicots, members of the clade containing Arabidopsis EXO70A1 usually exhibit high expression levels in sporophytic vegetative organs, with low or no expression in pollen, whereas a complementary pattern, i.e. high transcript level in pollen and lower to no expression in the sporophyte, is typical for the clade containing EXO70A2. Also, in the grasses, one of the two clades contains mostly genes with either ubiquitous or predominantly sporophytic expression, whereas most members of the other exhibit enhanced, though usually not exclusive, expression in the male gametophyte.
Preparation and Characterization of a Loss-of-Function exo70a2 Allele Using the CRISPR/Cas9 System To investigate the function of EXO70A2 in planta, we intended to use Arabidopsis mutants, but at the time of this study, T-DNA insertion null mutants in EXO70A2 (At5g52340) were not available. The publicly available GABI_824D06 line (exo70a2-1) with an insertion located in the 59 untranslated region exhibited EXO70A2 overexpression , and the FLAG_264F01 line could not be confirmed in our hands. We thus employed the CRISPR/Cas9 system (Wang et al., 2015) and targeted it specifically to the EXO70A2 gene to generate a loss-of-function exo70a2 mutant in the Arabidopsis ecotype Columbia-0 background. We obtained a mutant line (exo70a2-4) bearing a 13-bp insertion in the fifth exon (of 11 exons) of the EXO70A2 gene that caused a frameshift in the coding sequence, resulting in a premature stop codon (Supplemental Fig. S1A). The insertion likely affected the EXO70A2 mRNA stability or processing, because the EXO70A2 transcript level was reduced in the exo70a2-4 mutant to less than half the wild-type level as evaluated by semiquantitative reverse transcription PCR (RT-sqPCR) in flower material (Supplemental Fig. S1B).
Plants homozygous for the CRISPR-generated insertion (hereafter referred to as exo70a2) were then subjected to phenotype analysis and compared to their wild-type counterparts. To address whether EXO70A2 might play some role in the sporophyte despite its minimal expression in sporophytic tissues (Synek et al., 2006;Hruz et al., 2008;www.Genevestigator.com), we inspected the morphology of the mutant and wild-type plants and evaluated two basic morphological parameters. Primary root length of 7-d-old seedlings grown on vertical agar plates was comparable between exo70a2 and the wild type (Supplemental Fig. S1, C and D). Similarly, 5-week-old exo70a2 and wild-type plants showed no difference in plant height and general plant morphology (Supplemental Fig. S1, E and F). In conclusion, the sporophyte of exo70a2 homozygotes was not affected by the mutation, consistent with the notion that EXO70A2 is a key EXO70 isoform functioning in pollen.

The exo70a2 Mutant Shows a Pollen-Specific Transmission Defect
We next analyzed the transmission of the exo70a2 mutant allele through pollen. Based on PCR genotyping, the progeny of heterozygous exo70a2 mutant plants showed a significantly reduced portion of mutant homozygotes in comparison to the normal Mendelian ratio (only 15% of the tested plants were mutant homozygotes), pointing to a severe transmission defect of this mutant allele (Table 1).
To prove that the transmission defect was specific to males we performed reciprocal crossing of exo70a2 heterozygotes to wild-type plants and inspected the frequency of wild-type and heterozygous progeny ( Table 2). When exo70a2 heterozygous plants were used as pollen donors, only 17% of the progeny were heterozygous for the exo70a2 allele instead of the expected 50%. However, when they were used as pollen recipients, 54% of the progeny were heterozygous, indicating that the transmission defect was male specific.

Pollen Maturation and Germination Are Compromised in the exo70a2 Mutant
To investigate the nature of the pollen-specific transmission defect, we inspected pollen development, including pollen germination and pollen tube growth in exo70a2 and wild-type plants. Initially, we conducted an in vitro pollen germination assay on pollen samples harvested from homozygous exo70a2 mutants, heterozygous exo70a2 mutants, and wild-type controls. While the control pollen started to efficiently germinate within 2 h after imbibition, the pollen from exo70a2 homozygotes germinated no sooner than 8 h after the imbibition (Fig. 2, A and B). Even after 20 h, the exo70a2 pollen still showed 8-fold lower germination efficiency than wildtype pollen.
In order to resolve the pollen germination defect, we further analyzed pollen grain viability and several aspects of maturation. We observed a four-timeshigher percentage of nonviable pollen grains at the mature pollen stage in exo70a2 compared to the wild type using Alexander staining (wild type, 3.4%, n 5 179; exo70a2, 11.5%, n 5 157; P 5 0.004, x 2 test; Fig. 2C). An analysis of nuclear composition in the pollen using 49,6-diamino-phenylindole staining demonstrated that exo70a2 pollen grains could proceed through all developmental stages before germination (Supplemental Fig. S2A). However, a quantitative evaluation revealed that dehiscent anthers of exo70a2 contained a significantly higher portion of tricellular pollen grains at the expense of mature pollen grains protein sequences from representative land plants ranging from liverworts (Marchantia polymorpha), mosses (P. patens and Sphagnum fallax), lycophytes (Selaginella moelendorffii), gymnosperms (Gnetum sp., Picea sitchensis), the basal angiosperm Amborella trichopoda, and grasses (Brachypodium distachyon, O. sativa, Sorghum bicolor, and Zea mays) as representatives of the monocots, and both asterids (Solanum lycopersicum and Nicotiana tabacum) and multiple rosids (Vitis vinifera, Populus trichocarpa, Arabidopsis, Arabidopsis lyrata, Fragaria vesca, and Pyrus brettschneideri) as representatives of the eudicots. The sole EXO70 of the charophyte Klebsormidium flaccidum has been included as an outgroup (for a full list of sequences with accessions see Supplemental File S1). A consistent tree was also obtained by the maximum likelihood method (bootstrap values shown); full support for some branches is denoted by symbols. For selected angiosperm species, a heat map of transcript levels in pollen and sporophytic tissues, inferred from publicly available transcriptome data, is shown. ML, Maximum likelihood; n.a., not available.
Analysis of wild-type and exo70a2 mature pollen grains from open flowers using transmission electron microscopy revealed a higher density of endomembrane vesicles in cortical regions of exo70a2 pollen grains compared to the wild type (exo70a2, 840 6 81 vesicles•mm 22 [mean 6 SD], n 5 5; wild type, 538 6 116 vesicles mm 22 , n 5 5; P , 0.0014, Student's t test), directly indicating a defect in exocytosis in exo70a2 pollen grains (Supplemental Fig. S3) Since high concentrations of extracellular reactive oxygen species (ROS) were reported at the onset of the pollen germination process (Smirnova et al., 2014), we analyzed ROS production in wild-type and exo70a2 pollen grains 20 min after imbibition using histochemical staining of superoxide anion radicals by nitroblue tetrazolium (NBT). ROS levels were dramatically reduced in mutant pollen grains, indicating a different physiological status (Fig. 2,E and F).
Taken together, we conclude that EXO70A2 is involved in pollen grain maturation, thus representing an important function assigned to the exocyst complex in plants.

Pollen Tube Growth Is Compromised in the exo70a2 Mutant
To determine whether the mutation affects not only pollen germination, but also pollen tube elongation, we analyzed pollen tube lengths 20 h after imbibition in vitro and found that exo70a2 pollen tubes were dramatically shorter than wild-type ones (Figs. 2A and 3A). Subsequent analysis of growth dynamics of individual pollen tubes revealed that the average maximal growth rate of exo70a2 pollen tubes was six times lower than that of their wild-type counterparts (0.39 6 0.08 mm min 21 and 2.42 6 0.37 mm min 21 [mean 6 SD] for exo70a2 and the wild type, respectively; n . 12;, P , 0.0001, Student's t test). Interestingly, an inspection of the exo70a2 pollen tube elongation in vivo in self-pollinated pistils showed a less dramatic, yet significant, defect in pollen tube growth compared to the wild type (Fig. 3, B and C). This can be explained by the presence of chemical signals produced by the transmitting tract tissues in pistils that promote pollen tube growth in a dosedependent manner (Johnson and Preuss 2002;Vogler et al., 2014).
Taken together, the pollen-specific transmission defect of the exo70a2 mutant allele is due not only to impaired pollen germination, but partly to impaired pollen tube growth. The latter defect, i.e. very short and wide pollen tubes with compromised polarity, has been previously reported for other mutants in exocyst subunits (Cole et al., 2005;Hála et al., 2008).

Slow Growing exo70a2 Pollen Tubes Display Morphological Defects
In the in vitro pollen germination experiments, we noticed that exo70a2 pollen tubes had grown straight, without branching, but had an ;50% wider diameter than wild-type pollen tubes (Fig. 4, A and B). This morphological defect suggests that secretion is delocalized in the apex of exo70a2 pollen tubes, which is consistent with the canonical exocyst function in polarized exocytosis in tip growth.
Since the exocyst plays an important role in polarized exocytosis, which in turn is crucial for cell wall construction, we tested whether the pollen tube growth defect in exo70a2 might be due to altered cell wall biogenesis. To do so, we analyzed the distribution of dominant cell wall components in growing pollen tubes in vitro. Pectins showed normal distribution after propidium iodide staining (Fig. 4C). Callose and, especially, cellulose deposition visualized by aniline blue and calcofluor white staining, respectively, extended more toward the tip in exo70a2 pollen tubes (Fig. 4, D and E). This observation could be explained as a secondary effect of the much lower growth rate of mutant pollen tubes, because even wild-type pollen tubes exhibiting low growth rates, such as newly germinated or very mature pollen tubes, showed a similar cellulose deposition pattern (Supplemental Fig. S4). We  For further functional characterization of EXO70A2, we prepared a GFP-tagged variant of EXO70A2 expressed under its native promoter. The pEXO70A2::GFP:EXO70A2 construct was introduced into exo70a2 heterozygous plants to analyze EXO70A2 localization and functionality in mutant and wild-type pollen. The expression of pEXO70A2::GFP:EXO70A2 complemented the EXO70A2 disruption: the percentage of nonviable pollen grains in freshly open flowers was similar to that of the wild type (complemented line, 3.9%, n 5 231; wild type: 3.4%, n 5 179; P 5 0.772, x 2 test). The ability of exo70a2 pollen to germinate in vitro without any delay was restored (Fig. 5,A and B;Supplemental Fig. S5), the pollen tube growth defect was reverted to normal (Fig. 5C), and the pollen tube width was similar to wild-type tubes (6.20 6 0.45 mm 6.37 6 0.71 mm for the complemented line and the wild type, respectively [mean 6 SD]; n 5 32; P 5 0.257, Student's t test). These observations confirm that the reported defects in the exo70a2 pollen were due to the disruption of the EXO70A2 gene.
While seedlings and mature sporophytic organs exhibited no fluorescence signal attributed to GFP:EXO70A2 (not shown), we clearly detected GFP:EXO70A2 in pollen grains and pollen tube tips (Fig. 6). GFP:EXO70A2 was localized mostly in the nucleolus in unicellular pollen and in both vegetativeand generative-cell nuclei in bicellular pollen. In tricellular pollen, it strongly accumulated in the whole cytoplasm up to the PM and periphery of the vegetative nucleus, with some remnants inside the nucleus, and in mature pollen grains, it was weakly present in the cytoplasm, enriched around the vegetative-cell nucleus and sperm cells, but excluded from the lumen of all nuclei (Fig. 6A). The obviously higher signal intensity of GFP:EXO70A2 at the tricellular stage corresponds to the maximum level of EXO70A2 mRNA expression across pollen developmental stages based on microarray data . Importantly, in complemented exo70a2 pollen tubes, GFP:EXO70A2 displayed a highly polarized localization along the PM in the apical domain of growing pollen tube tips, with a minor portion in the cytoplasm (Fig. 6B), which is a typical localization pattern for exocyst subunits in pollen tubes (Hála et al., 2008;Synek et al., 2017). In the wild-type background, however, the polar PM localization was absent and GFP:EXO70A2 was distributed entirely in the cytoplasm of pollen tubes (Fig. 6C), which could be explained by a competition of the native and GFPtagged EXO70A2 variants for binding to the exocyst complex, as observed previously in tobacco (Nicotiana  Fig. 2A). More than 100 pollen tubes were evaluated for wild-type and heterozygous plants, and 50 pollen tubes for the exo70a2 mutant. Pollen was collected from five different plants for each genotype, with three biological replicates. B, Aniline blue staining for visualization of callose in pollen tubes in representative self-pollinated pistils from a wild-type and a homozygous exo70a2 plant. Dashed lines indicate the longest pollen tubes. C, Quantification of the maximal pollen tube length based on aniline blue staining. At least 30 pistils of each genotype were evaluated, with two biological replicates. Error bars represent the mean 6 SD. Asterisks indicate statistical difference using Student's t test (P , 0.001). tabacum) pollen tubes , and also in mammalian cells (Matern et al., 2001).

EXO70A2 Can Rescue the exo70a1 Mutation in the Sporophyte
Next, we were interested in the degree of functional specialization versus redundancy of EXO70A1 and EXO70A2 paralogs. In order to address this question experimentally, we expressed EXO70A2 and EXO70A1, respectively, under control of the EXO70A1 promoter in the exo70a1 mutant background and compared their localization and functionality in the sporophyte. The exo70a1 mutant was earlier shown to exhibit pleiotropic morphological defects, including retarded polar growth of root hairs, loss of apical dominance, dwarfish stature, and sterility (Synek et al., 2006). Moreover, GFP:EXO70A1 fully complemented the exo70a1 loss-of-function allele and localized at the PM, with special enrichment at the outer lateral domain in root epidermal cells Fendrych et al., 2013).
We found that the expression of pEXO70A1::GFP:-EXO70A2 rescued the morphological defects of exo70a1 mutant plants similarly to pEXO70A1::GFP:EXO70A1, as demonstrated by measurements of plant height and overall morphology (Fig. 7, A and B). Importantly, the subcellular localization of GFP:EXO70A2 in root epidermal cells was indistinguishable from that of GFP:EXO70A1 (Fig. 7C). We conclude that EXO70A1 and EXO70A2 paralogs have a potential to be functionally redundant in the sporophyte.
Furthermore, the EXO70A2 disruption had an additive effect on the disruption of EXO70A1. In contrast to small exo70a1 mutants, seedlings of the exo70a1 exo70a2 double mutant were even smaller short-living dwarf plants that exhibited significantly shorter roots and shoots, characterized by asymmetric, reduced, or completely missing cotyledons (Fig. 7, D and E; Supplemental Fig. S6). Subsequent analysis of EXO70A2 expression in wild-type, exo70a1, and exo70a2 seedlings using RT-sqPCR revealed that EXO70A2 is weakly expressed in the sporophyte and its expression is upregulated in the exo70a1 knockout when compared to wild-type seedlings (Supplemental Fig. S7). The PCR product was sequenced to prove its EXO70A2 identity. Collectively, the comparison of single-and doublemutant phenotypes and the detection of EXO70A2 in seedlings suggest that EXO70A2 is expressed even in the sporophyte, in addition to its dominant role in the male gametophyte, and is able to partly substitute for the EXO70A1 function.

EXO70A1 Cannot Substitute for EXO70A2 in the Pollen
In parallel with the previous experiment, we tested whether EXO70A1 could provide a similar function to EXO70A2 in pollen when ectopically expressed under the control of the EXO70A2 promoter in the exo70a2 background. pEXO70A2::GFP:EXO70A1 was detected in a pattern similar to that observed for pEXO70A2::GFP:EXO70A2 in exo70a2 pollen grains (Fig. 8A). However, pollen of these transgenic plants was unable to germinate, although we analyzed six plants originating from two independent transformations done using the same cloning strategy and materials as for the previous constructs (Fig. 8B). We conclude that when EXO70A2 is disrupted, EXO70A1 cannot substitute for its function in pollen, despite the functional redundancy of these two isoforms observed in the sporophyte.
To explain this discrepancy, we compared structural models of EXO70A1 and EXO70A2 based on crystallographic data generated from yeast (Saccharomyces cerevisiae) Exo70 and Arabidopsis EXO70A1 (Dong et al., 2005;Hamburger et al., 2006;Zhang et al., 2016;Pleskot et al., 2015;Mei et al., 2018). We noticed substantial differences in the amino acid composition in a medial plant-specific loop and on the side of the EXO70A2 molecule facing out of the PM (Supplemental Fig. S8). The variability was manifested by a different electrostatic potential on the EXO70A2 surface in the loop and C terminus. This analysis suggests that EXO70A1 versus EXO70A2 may interact with specific proteins in the sporophyte or male gametophyte, respectively, that modulate their function.

DISCUSSION
Proper pollen development, efficient pollen germination, and pollen tube growth are essential processes for successful plant fertilization. These processes include cell expansion and cell wall maturation, which require highly efficient secretory machinery, delivering cell wall components and PM-associated materials to specific sites at the PM. To sustain such demands, polarized secretion is regulated by a number of molecular regulators, including the vesicle-tethering exocyst complex. The crucial role of several core exocyst subunits in pollen germination and pollen tube tip growth was previously described (Cole et al., 2005;Hála et al., 2008;Bloch et al., 2016). However, the main EXO70 isoform contributing to the conventional secretory function of the exocyst complex in pollen has not been characterized until now.
We previously described how EXO70.2 clade members EXO70C2 and EXO70C1 are not stable subunits of the exocyst complex , leaving EXO70A2 as the best candidate for the main housekeeping EXO70 isoform in pollen. First, EXO70A2 is the closest paralog to EXO70A1-the main EXO70 isoform in the Arabidopsis sporophyte-and the two share 72% sequence identity at the protein level. Second, similar to EXO70A1, EXO70A2 physically interacted with three other exocyst subunits (SEC3a, SEC10b, and EXO84b; Hála et al., 2008;Fendrych et al., 2010;Synek et al., 2017). Third, the expression patterns of EXO70A1 is sporophyte specific, whereas that of EXO70A2 is pollen specific. This is in agreement with the hypothesis that after whole-genome duplication in plants, genes directly linked to cell polarity tend to diverge in expression pattern to sporophyte specific and to those enriched in tip-growing cells, especially in the male gametophyte (De Smet et al., 2017). In general, about 10% to 40% of Arabidopsis genes are selectively expressed in pollen, and these genes are enriched in signaling, vesicle trafficking, the cytoskeleton, and membrane transport (Becker et al., 2003;Honys and Twell, 2003). Importantly, EXO70A2 still retains the capacity to function as an exocyst subunit in sporophytic cells, whereas EXO70A1 is unable to substitute for EXO70A2 function in pollen, possibly due to its inability to bind EXO70A2-specific protein interactors in the male gametophyte.
In the course of the extraordinary diversification and functional specialization of the large EXO70 family in angiosperms (Cvrčková et al., 2012), the EXO70A duplication to sporophyte-and male gametophytespecific paralogs occurred independently in monocots and eudicots. In some plant taxonomic groups, each further duplicated, generating, for example, EXO70A2 and EXO70A3 in Brassicaceae (Cvrčková et al., 2012). An analogous situation, i.e. the presence of pollenexpressed and sporophyte-expressed paralogs, has been documented in the case of SEC3, SEC15, and EXO84 exocyst subunits in Arabidopsis (Bloch et al., 2016;Synek et al., 2017; K. Batystová, M. Klejchová, E. Janková Drdová, L. Synek, P. Sabol, M. Potocký, V.Žárský, and M. Hála, unpublished data). Analagous to our observation that artificial EXO70A2 expression in the sporophyte could complement the sporophytic defects of exo70a1 mutants, SEC15a, the main SEC15 isoform in Arabidopsis pollen, could substitute for the function of SEC15b in the sporophyte, although SEC15a and SEC15b share a mere 47% sequence identity at the protein level (K. Batystová, M. Klejchová, E. Janková Drdová, L. Synek, P. Sabol, M. Potocký, V.Žárský and M. Hála, unpublished data). In contrast, Arabidopsis SEC10a and SEC10b represent a very recent duplication, with no diversification in their expression or function (Vukašinović et al., 2014).
Based on the phenotypic differences observed in exo70a2 mutants, we documented that EXO70A2 participates in pollen maturation, germination, and pollen tube growth. In line with this, other recently reported exo70a2 point mutation alleles show defects in germination efficiency and pollen tube growth (Beuder et al., 2020). However, in contrast to that report, our data imply that the primary defect in exo70a2 mutants may be related to its function in pollen maturation and that the germination defect might in fact be a combination of both compromised pollen maturation and impaired tip targeting of the exocytosis pathway. Similarly, mutants in two other exocyst subunits, sec3a and sec8, also exhibited severe defects in germination efficiency (Cole  Bloch et al., 2016). A higher percentage of nonviable and bursting pollen grains and delayed germination in exo70a2 could be explained by mislocalization or inefficient delivery of some cell wallmodifying enzymes (or cell wall components) whose trafficking during pollen maturation and germination depends on the EXO70A2-containing exocyst complex. Indeed, we observed accumulation of endomembrane vesicles in cortical regions of exo70a2 pollen grains. In agreement, it has been documented that proper structure and chemical characteristics of the cell wall are essential for efficient pollen germination and pollen tube growth (Chebli et al., 2012;Leroux et al., 2015;MacAlister et al., 2016). For example, the Arabidopsis mutant in pollen-specific pectin methylesterase (PME48) showed delayed pollen grain germination and bursting of pollen tubes (Leroux et al., 2015). Likewise, triple and quadruple mutants in pollen-expressed Leurich repeat extensins (LRX8-LRX11) showed a high proportion of ungerminated and burst pollen grains and defective pollen tubes . Also, the reduction of pectic arabinan side chains, caused by ectopic expression of endo-a-1,5-L-arabinanase (eGARA), resulted in collapsed pollen grains with altered intine during pollen maturation in potato (Solanum tuberosum; Cankar et al., 2014).
One of the crucial regulatory modules in pollen germination and pollen tube growth is ROS production (Potocký et al., 2007(Potocký et al., , 2012Smirnova et al., 2014). Two Arabidopsis pollen-specific NADPH oxidases localized at the PM, RBOHH and RBOHJ, were necessary for ROS accumulation in the pollen grain cell wall upon pollination and later in pollen tube tips. Double mutants in these genes exhibited a much lower level of ROS, resulting in collapsing pollen tubes (Lassig et al., 2014;Kaya et al., 2015). Interestingly, data of Smirnova et al. (2014) suggest that targeted production of extracellular ROS is required for changes in mechanical properties of the intine in pollen grains underlying efficient germination. Since we observed a significant decrease in ROS production in exo70a2 germinating pollen grains, we suggest that the pollen exocyst may function in the delivery of NADPH oxidases to specific sites at the PM during germination.
Following pollen germination, rapidly elongating pollen tubes require precisely regulated secretory machinery to support their intensive tip growth. Similar to exo70a2 in this study, mutants in several core exocyst subunits (sec5a/b, sec6, sec8, and sec15a) generate extremely short and wide pollen tubes with drastically reduced fertilization capacity (Hála et al., 2008). Reduced transmission efficiency and pollen tube elongation of exo70a2 mutants were observed also by Beuder et al. (2020). The pollen tube phenotype is most likely caused by an inefficient targeting of secretory vesicles, possibly carrying some specific cargo, to the growing tip. For example, mutants in Hyp O-arabinosyltransferases HPAT1 and HPAT3, involved in modification of cell wall-associated extensins (MacAlister et al., 2016), and pectin methylesterases PPME1 (Tian et al., 2006) and VANGUARD1 (Jiang et al., 2005) exhibited a pollen tube elongation defect similar to that observed in exo70a2. Cell wall and pollen tube morphology defects of the hpat1 hpat3 double mutant were partly suppressed by disruption of EXO70A2, and also SEC15a, further indicating the involvement of EXO70A2 and possibly the whole exocyst complex, in targeted secretion in growing pollen tube tips (Beuder et al., 2020). However, our data clearly show that the EXO70A2 function is essential for efficient pollen maturation. The PM localization of EXO70A2 in growing pollen tube tips (also observed in Beuder et al., 2020) was identical to that of core exocyst subunits SEC8, SEC10a, and SEC15a K. Batystová, M. Klejchová, E. Janková Drdová, L. Synek, P. Sabol, M. Potocký, V.Žárský and M. Hála, unpublished data). At the molecular level, EXO70A2 likely mediates the exocyst interaction with the PM, similar to Exo70 in yeast and mammalian cells (Boyd et al., 2004;He et al., 2007;Pleskot et al., 2015). This notion is supported by the observation that the N-terminally truncated SEC3a subunit in Arabidopsis pollen was unable to interact with the PM but still localized to the apical PM in growing pollen tube tips as a part of the complex (Bloch et al., 2016). However, SEC3, known to mediate interaction of the Arabidopsis exocyst with active ROP GTPases via the ICR1 adaptor protein (Lavy et al., 2007), might perhaps be at least partly functionally redundant in exocyst PM targeting with the EXO70A2 function. ROP GTPases and their regulators (such as the REN1 GAP) are known to be localized at the pollen tube tip, similar to EXO70A2, and are crucial for cell polarity establishment and maintenance (Hwang et al., 2008;Li et al., 2008).
Although the transmission efficiency of the exo70a2 mutant allele was heavily impaired, it was not as strong as in loss-of-function mutants (sec6, sec8, and sec15a) in the core exocyst subunits (Hála et al., 2008). This indicates that some other EXO70 isoform(s) could provide a function partially overlapping with that of EXO70A2, with possible candidates including EXO70H3, EXO70H5, and EXO70H6. However, the abundance of these isoforms in the pollen proteome is most likely very low , and their relevance thus would have to be proved experimentally. Alternatively, EXO70A1 might be activated when EXO70A2 is disrupted, despite the fact that EXO70A1 is normally not expressed in pollen (www.Genevestigator.com; Hruz et al., 2008;Grobei et al., 2009). We propose that the other six EXO70 isoforms expressed in pollen adopted specific functions showing no or only partial functional overlap with that of EXO70A2. In addition, their activities might be restricted to certain stages of pollen development or specific endomembrane domains.
In summary, we conclude that among seven EXO70 isoforms expressed in the Arabidopsis male gametophyte, EXO70A2 is the main isoform, contributing to the conventional exocyst function in exocytosis, like its sibling, EXO70A1, in the sporophyte. However, EXO70A2 is important not only for pollen germination and highly polarized pollen tube growth, but also for pollen grain maturation. This indicates that the exocyst plays a role in functional continuity between pollen maturation and the following processes, pollen germination and pollen tube growth. Despite the deep evolutionary split in eudicots, EXO70A2 still retains the ability to fully substitute for the EXO70A1 function in the sporophyte, whereas EXO70A1 lost the capacity to function in pollen development. The duplication of EXO70A genes to sporophyte-and male gametophytespecific paralogs occurred independently in monocots and eudicots, suggesting convergent functional evolution of pollen-specific versions of the exocyst complex in angiosperms-an interesting challenge for future explorations.

Phylogenetic and Expression Analysis
To perform the phylogenetic analysis, a set of EXO70.1 clade sequences from our previous studies (Cvrčková et al., 2012;Rawat et al., 2017) has been updated according to their newest database status, and homologs from additional plant species were obtained by BLASTP searches of the GenBank, Phytozome (Goodstein et al., 2012), and GDR (www.rosaceae.org) databases using Arabidopsis (Arabidopsis thaliana) EXO70A1 (At5g03540) and EXO70A2 (At5g52340) sequences as a query. The full list of sequences is provided in Supplemental File S1. Protein sequence alignment was constructed using the MAFFT E-INS-I algorithm (Katoh and Standley, 2013) in Jalview software (Waterhouse et al., 2009). Gaps and nonconserved regions were then manually deleted from the alignment, giving a matrix of 56 taxa and 546 positions. Bayesian phylogeny inference was performed using MrBayes software (Ronquist et al., 2012) with a WAG amino acid model, where the analysis was performed in four runs with four chains and 500,000 generations, and trees were sampled every 100 generations. All four runs asymptotically approached the same stationarity after the first 125,000 generations, which were omitted from the final analysis. Maximum-likelihood phylogeny was performed in Phyml software (Guindon et al., 2010) using the LG matrix g-corrected for among-site rate variation with four rate site categories plus a category for invariable sites, with all parameters estimated from the data model to build the phylogenetic tree. Bootstrap analysis (500 replicates) was performed to estimate the probability of maximumlikelihood tree topology.
Analysis of the EXO70.1 paralog expression in male gametophytic and sporophytic tissues was performed using the CoNekT (Proost and Mutwil, 2018) and GEO (Barrett et al., 2013) tools. Transcript abundance of tobacco EXO70.1 isoforms in pollen was assessed by reanalyzing RNA-sequencing data generated by Conze et al. (2017) according to Ghosh and Chan (2016). We used transcripts per kilobase million-based normalization, because it could be used for gene count comparisons both within a sample and between samples of the same sample group (Abrams et al., 2019). Briefly, raw data reads were analyzed using fastQC and mapped to a reference gene-set (TN90 cultivar) using tophat. Cufflinks was used to assemble and reconstruct the transcriptome and calculate transcripts per kilobase million expression values. Additional anther/pollen RNA-sequencing data for sorghum (Sorghum bicolor), Arabidopsis, poplar (Populus spp.), strawberry (Fragaria spp.), and pear (Pyrus spp.) were obtained from Davidson et al. (2012), Loraine et al. (2013), Zhao et al. (2016), Zhou et al. (2016), andHollender et al. (2014), respectively.
Seeds target single-guide RNAs were designed. Sequences of the CRISPR target sites were carefully checked by BLAST against all other EXO70 paralogs in the Arabidopsis genome to exclude unintended targeting of additional EXO70 paralogs (target site 1, AGCTAAAATATTGAG; target site 2, GGA-TACTCGAGCTG). Flowering Arabidopsis plants of the Columbia-0 ecotype (NASC collection) were transformed using the Agrobacterium tumefaciensmediated floral dip method (Clough and Bent, 1998). After selection on hygromycin, mutant lines were identified by sequencing of PCR products obtained from the targeted part of the EXO70A2 gene (At5g52340). A mutant line having a 13-bp insertion in the expected site (...TCGAGCTGCGGTttcatgcga ttttGTTGGAACAGAG...), causing a premature stop codon due to a frameshift, was further characterized. The genotype was then routinely analyzed by PCR genotyping, where two primers, one from each pair for the wild type and mutant allele, were designed over the CRISPR-modified site (Supplemental Fig.  S1A; Supplemental Table S1).

Semiquantitative RT-PCR Analysis of the EXO70A2 Transcription
To determine the EXO70A2 transcript level in the CRISPR-modified mutant line, total RNA was extracted from 100 mg of apical parts of primary inflorescences of exo70a2 and wild-type plants using the RNeasy kit (Qiagen). Prior to reverse transcription, samples were treated by DNase I (New England Biolabs). The complementary DNA (cDNA) was synthesized using 2 mg of total RNA, oligo(dT) primers and Transcriptor High Fidelity cDNA Synthesis kit (Roche). Transcript abundance was assayed by RT-sqPCR for two gene regions, upstream and downstream from the CRISPR-generated insertion (see Supplemental Fig.  S1), using EXO70A2-specific primers (Supplemental Table S1); the primers were checked for potential interference with EXO70A1 and EXO70A3 paralogs. ACTIN7-specific primers were used as a quantitative control (Supplemental Table  S1). The optimal number of PCR cycles was determined empirically.
For determination of the EXO70A2 transcript level in the sporophyte, the procedure was similar, with the following differences: total RNA was extracted from 100 mg of 20-d-old seedlings of the wild type, exo70a1, and exo70a2. Primers were specific for the downstream part of the EXO70A2 cDNA (Supplemental Table S1).

Analysis of Pollen Development
Pollen grains at different developmental stages were collected from a series of flower buds ending with a freshly opened flower and used for further analyses of pollen grain viability, maturation, and nuclear composition.
To inspect pollen grain viability, we used Alexander staining (Alexander, 1969) with the following modifications: 103 dilution, 15 min incubation time at room temperature. Images were captured using a Nikon Eclipse 90i microscope with Plan Apo 103/0.45 objective, differential interference contrast (DIC) optics, and a Zyla sCMOS camera (Andor). Fully developed viable and nonviable pollen grains were scored. Underdeveloped collapsed grains were excluded from the analysis.
To analyze the nuclear composition, pollen grains were stained with 49,6-diamino-2-phenylindole at a final concentration of 4 mg mL 21 in pollen isolation buffer (100 mM Na 3 PO 4 [pH 7.5]; 1 mM EDTA; and 0.1% [v/v] Triton X-100) as described by Backues et al. (2010). Images were captured using a Zeiss Axio Imager 2 microscope with an EC Plan-Neofluar 403/0.75 objective, a filter set for observation of blue fluorescence, DIC optics, and a Zeiss Axiocam 506 Color camera.
For analysis of the nuclear composition, pollen grains expressing GFP:EXO70A2 were stained with Hoechst 33258 at a final concentration of 1 mg mL 21 in liquid pollen germination medium. After 30 min, pollen grains were imaged using a Zeiss LSM 880 confocal laser scanning microscope with Plan-Apochromat 403/1.2 WI objective.
Pollen from 10 open flowers was harvested from five plants per genotype and extracted in 1 mL of fresh liquid pollen germination medium by three times repeated vortexing and centrifugation steps (1 min vortexing, 3 min centrifugation at 3,000 rpm). Pollen pellets were resuspended in 15 mL of Ruthenium red solution in distilled water (final concentration 0.14 mg mL 21 ). Images were taken using a Nikon Eclipse 90i microscope with Plan Apo 43/0.2 and Plan Apo 103/0.45 objectives.

Transmission Electron Microscopy
For transmission electron microscopy, pollen suspensions from open flowers of the wild type and exo70a2 were fixed for 24 h in 2.5% (v/v) glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2) and postfixed in 2% (w/v) OsO 4 in the same buffer. Samples were then dehydrated through an ascending ethanol and acetone series and embedded in Araldite-Poly/Bed 812 mixture. Thin sections (70 nm) were cut on a Reichert-Jung Ultracut E ultramicrotome and stained using uranyl acetate and lead citrate. Sections were examined and photographed using a JEOL JEM-1011 electron microscope.

Germination of Arabidopsis Pollen In Vitro
Pollen was germinated on a thin layer of semisolid pollen germination medium (10% [w/v] Suc, 1.5% [w/v] low-melting-point agarose [Agarose-Universal, peqGOLD, VWR], 0.01% [w/v] H 3 BO 3 , 5 mM CaCl 2 , 5 mM KCl, and 1 mM MgSO 4 ) in a chambered Lab-Tek II coverglass (Thermo Scientific). Pollen grains from fully opened flowers were spread onto the medium layer and chambers were closed and placed into standard plant growth conditions (see above).
To evaluate germination efficiency, pollen grains were scored as germinated when a pollen tube length reached at least one-half of the pollen grain diameter. For each genotype, 10 samples originating from five different plants were observed. Pictures of germinated pollen were taken using a Nikon Eclipse 90i microscope with a Plan Apo 103/0.45 objective. Pollen tube lengths were measured in ImageJ and histograms were generated in Microsoft Excel.

ROS Staining
Production of superoxide was determined by its ability to reduce NBT to formazan precipitate (Rossetti and Bonatti, 2001). Pollen grains were spread on semisolid pollen germination medium (see above). After 20 min of imbibition, they were overlaid with 20-mL drops of NBT diluted in liquid pollen germination medium to a final concentration 2 mg mL 21 , incubated for 10 min, and immediately imaged using a Nikon Eclipse 90i microscope with a Plan Apo 203/0.75 objective, excluding DIC filters, and maximum field aperture opening. NBT signal intensity was measured in ImageJ. Values were normalized to that of nonstained pollen grains and related to maximal staining intensity value across all samples. The experiment was repeated in three replicas with similar results.

Pollen Tube Morphology and Growth Rate
Details of pollen tube morphology were captured using a Zeiss Axio Imager 2 microscope with an EC Plan-Neofluar 403/0.75 objective at 3 and 16 h after imbibition for the wild type and exo70a2, respectively. The pollen tube growth rate was analyzed using the same microscope but with an EC Plan-Neofluar 203/0.5 objective. Time-lapse photographs at 2-min intervals were recorded for 30 min for the wild type (starting 3 h after imbibition) and 90 min for exo70a2 (starting 16 h after imbibition). Only pollen tubes longer than twice the pollen grain diameter were evaluated, because they already had reached the maximal growth rate.

Callose Visualization in Arabidopsis Pistils
Self-pollinated pistils (at least 30 for each genotype) from fully opened flowers were collected and stained with aniline blue according to Mori et al. (2006) and imaged using a Nikon Eclipse 90i microscope with a Plan Apo 43/ 0.2 objective and a Zyla sCMOS camera (Andor). Length of the longest pollen tube in every pistil was measured, and recorded values were evaluated statistically using Student's t test.

Fluorescent Staining of Cell Wall Components
Propidium iodide, Calcofluor White, or Aniline Blue was diluted in liquid germination medium and gently applied onto pollen germinated in vitro immediately before imaging. Images were captured using a Zeiss LSM 880 confocal laser scanning microscope with a Plan-Apochromat 103/0.45, Plan-Apochromat 203/0.8, C-Apochromat 403/1.2 WI, and C-Apochromat 633/ 1.2 WI objectives. Working concentrations, excitation laser wavelengths, and range of recorded emission wavelengths were as follows: 30 mM, 514 nm, and 566 to 719 nm for propidium iodide; 1 mg mL 21 , 405 nm, and 410 to 523 nm for Calcouor White; and 0.001% (w/v), 405 nm, and 410 to 535 nm for decolorized Aniline Blue. Pollen was collected from five different plants for each genotype, and the experiment was repeated in two biological replicates.
For cloning of pEXO70A1::GFP:EXO70A2 and pEXO70A1::GFP:EXO70A1, the EXO70A1 promoter (1 kb upstream from the start codon) was subcloned into pDONR P4-P1r. The EXO70A1 was subcloned into pDONR P2R-P3. Then, multisite reactions were performed to assemble the EXO70A1 promoter, GFP, and EXO70A1 or EXO70A2 into the destination vector pB7m34GW. The insertions in pDONR vectors were sequenced using M13 primers. The final constructs were sequenced using M13 primers and two GFP primers (see Supplemental Table S1).
The subcellular localization of GFP:EXO70A2 in pollen grains, pollen tubes, and roots was performed using a Zeiss LSM 880 confocal laser scanning microscope equipped with a C-Apochromat 633/1.2 WI. The fluorophore was excitation with a 488-nm laser, and emitted fluorescence was recorded at 493 to 535 nm. The pollen was germinated as described above and carefully transferred to the microscope 2 h after imbibition.

Accession Numbers
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers 835310 (EXO70A2; protein NP_200047.3) and 831809 (EXO70A1; protein NP_195974.2).

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Characterization of the exo70a2 mutant line generated using the CRISPR/Cas9 system.
Supplemental Figure S2. Development of wild-type and exo70a2 pollen.
Supplemental Figure S3. Electron microscopic images of cortical regions in wild-type and exo70a2 pollen grains.
Supplemental Figure S4. Visualization of cellulose deposition in slowly growing wild-type pollen tubes in vitro.
Supplemental Figure S5. Pollen germination efficiency in the wild type and complemented exo70a2 mutants at different time points during in vitro germination.
Supplemental Figure S6. Morphological parameters of the exo70a1 exo70a2 double mutant in comparison to exo70a1 and exo70a2 single mutants and wild-type plants.
Supplemental Figure S7. Analysis of EXO70A2 expression in the sporophyte.
Supplemental Figure S8. Comparison of structural models of EXO70A1 and EXO70A2.
Supplemental Table S1. List of primers used in this study.
Supplemental File S1. EXO70 sequences used for the phylogenetic analysis.

ACKNOWLEDGMENTS
We thank Marta Cadyová for technical support, Said Hafidh for technical advice, and Juraj Sekereš for critical reading of the manuscript. Microscopy was performed in the Laboratory of Confocal and Fluorescence Microscopy at the Charles University and in the microscopic department in the Institute of Experimental Botany.