The canonical α-SNAP is essential for gametophytic development in Arabidopsis

Summary

The development of male and female gametophytes is a pre-requisite for successful reproduction of angiosperms. Factors mediating vesicular trafficking are among the key regulators controlling gametophytic development. Fusion between vesicles and target membranes requires the assembly of a fusogenic soluble N-ethylmaleimide sensitive factor attachment protein receptors (SNAREs) complex, whose disassembly in turn ensures the recycle of individual SNARE components. The disassembly of post-fusion SNARE complexes is controlled by the AAA⁺ ATPase N-ethylmaleimide-sensitive factor (Sec18/NSF) and soluble NSF attachment protein (Sec17/α-SNAP) in yeast and metazoans. Although non-canonical α-SNAPs have been functionally characterized in soybeans, the biological function of canonical α-SNAPs has yet to be demonstrated in plants. We report here that the canonical α-SNAP in Arabidopsis is essential for male and female gametophytic development. Functional loss of the canonical α-SNAP in Arabidopsis results in gametophytic lethality by arresting the first mitosis during gametogenesis. We further show that Arabidopsis α-SNAP encodes two isoforms due to alternative splicing. Both isoforms interact with the Arabidopsis homolog of NSF whereas have distinct subcellular localizations. The presence of similar alternative splicing of human α-SNAP indicates that functional distinction of two α-SNAP isoforms is evolutionarily conserved.

Keywords:

Anthers – Arabidopsis thaliana – Confocal laser microscopy – DAPI staining – Ovules – Plant genomics – Pollen – Embryo sac

Introduction

The development of male and female gametophytes is a pre-requisite for successful reproduction of angiosperms. In angiosperms, megagametogenesis [1] and microgametogenesis [2, 3] produce female and male gametophytes, respectively. During megagametogenesis, meiosis of a megaspore mother cell produces four megaspores, among which only one survives as functional megaspore (FM). FM undergoes three rounds of mitosis and cellularization to develop into an embryo sac, i.e. the female gametophyte [1]. During microgametogenesis, meiosis of a microspore mother cell gives rise to a tetrad of microspores. After being released from the tetrad, each microspore goes through an asymmetric cell division, referred to as pollen mitosis I (PMI), to produce a bicellular microspore containing a generative cell and a vegetative nucleus. The generative cell then undergoes another mitotic event, called pollen mitosis II (PMII), to produce two sperm cells enclosed in pollen together with the vegetative nucleus [2, 3].

Regulators of mitosis [4–7], of ribosomal biogenesis [8–12], and of endomembrane integrity [13–19] are major factors controlling gametogenesis. Soluble N-ethylmaleimide sensitive factor attachment protein receptors (SNAREs) are coiled-coil domain proteins regulating vesicular fusion [20, 21] between two membranous compartments, often vesicles and organelles within the endomembrane system. A fusogenic SNARE complex consists of four SNARE proteins [20, 21]. Mutations of SNAREs or their interacting partners often compromise gametophytic transmission [17–19, 22]. Indeed, functional loss of Arabidopsis YKT61, an R-SNARE protein, resulted in complete male and female gametophytic lethality [23], suggesting that SNARE-mediated membrane fusion is essential for gametophytic viability.

Vesicle-target membrane fusion not only depends on the assembly of tetrameric SNARE complex but also its disassembly so that the components of post-fusion SNARE complexes can be recycled [24]. Studies in yeast and metazoans demonstrated that the disassembly of post-fusion SNARE complexes is controlled by the AAA⁺ ATPase N-ethylmaleimide-sensitive factor (NSF/Sec18) and soluble NSF attachment protein (α-SNAP/Sec17), which perform ATP-dependent disassembly of cis-SNARE complexes, liberating SNAREs for subsequent assembly of trans-complexes for fusion [25, 26]. In addition to being the partner for NSF, α-SNAP performs a regulatory role in SNARE disassembly [24] or moonlights in other cellular processes [27, 28].

Recent studies in soybean showed that a naturally occurring, truncated α-SNAP allele, i.e. non-canonical α-SNAP, suppresses parasitic nematode infection [29–31]. The non-canonical α-SNAP may be derived from neofunctionalization after genome duplication [31, 32]. The non-canonical α-SNAP did not interact with the NSF homolog in soybean and its enhanced expression depleted the abundance of SNARE-recycling 20S complexes [29, 30]. Naturally occurring, truncated alleles of α-SNAP confer resistance against nematodes in soybean while the expression of a canonical α-SNAP counteracted the cytotoxicity of resistance-type Rhg1 α-SNAP [31, 32]. These results suggested that the non-canonical α-SNAP interferes with the role of the canonical α-SNAP in SNARE disassembly. However, the biological function of canonical α-SNAPs have yet to be demonstrated in plants.

We report here that the canonical α-SNAP in the Arabidopsis genome, designated ASNAP, is essential for male and female gametophytic development. By CRISPR/Cas9-mediated genomic editing, we generated and characterized asnap mutants. Functional loss of ASNAP resulted in gametophytic lethality such that both male and female gametophytes could not be transmitted. Specifically, functional loss of ASNAP caused mitotic arrest of unicellular microspores and of functional megaspores (FM), suggesting that ASNAP is essential for mitotic cell cycle progression during gametophytic development. We show that Arabidopsis ASNAP encodes two isoforms due to alternative splicing, both of which interact with the Arabidopsis NSF. The presence of similar alternative splicing of human α-SNAP indicates that functional distinction of two α-SNAP isoforms is evolutionarily conserved.

Results

Arabidopsis encodes one canonical α-SNAP

By sequence alignment, one canonical α-SNAP whose protein products contain N-terminal, central, and C-terminal domains similar to animal α-SNAP homologs are encoded in the Arabidopsis genome. To determine its expression pattern, we generated ASNAPg:GUS transgenic plants expressing genomic-GUS translational fusion of ASNAP. By histochemical GUS staining, we detected GUS signals in various tissues and developmental stages, including seedlings, leaves, roots, reproductive organs, trichomes, root hairs as well as pollen tubes (Fig 1). The constitutive expression of ASNAP is consistent with its role as a canonical α-SNAP.

Generation and characterization of asnap mutants

Because no valid T-DNA insertion lines were available for ASNAP from all stock centers, we used the genome-editing technology CRISPR/Cas9 [33, 34] to generate asnap mutants. We transformed Cas9-ASNAP driven by an egg cell-specific promoter [35] and screened its transformants for the editing of the ASNAP genomic locus. We identified three allelic mutations of ASNAP, in which nucleotide insertions resulted in pre-mature stop codons in the coding sequence of ASNAP (Fig 2A and 2B). Because asnap-1 is an allele repeatedly obtained, most experiments including the complementation assays were performed with asnap-1/+.

All three asnap mutant alleles were only obtained in their heterozygous forms, i.e. asnap-1/+, asnap-2/+, and asnap-3/+. Silique analysis showed that around 50% ovules were tiny white and wrinkled in the self-fertilized asnap/+ plants (Fig 2C–2G), indicating reduced fertility. To determine what have caused the seed set reduction, we examined asnap-1/+ pistils pollinated with wild-type pollen at 24 hours after pollination (HAP) by confocal laser scanning microscopy (CLSM) and whole-mount ovule clearing. At 24 HAP, wild-type ovules contained elongating zygotes or early embryos and peripheral endosperms (Fig 2H and 2J), indicating the completion of fertilization. By contrast, half of the asnap-1/+ ovules contained a single nucleus (Fig 2I) with no detectable peripheral endosperms or embryos (Fig 2I and 2K). These results suggested that the 50% white and wrinkled ovules in the heterozygous asnap/+ plants were not fertilized. Reciprocal crosses and seed set assays between asnap-1/+ and wild type showed that the reduced seed sets of asnap-1/+ were due to female gametophytic defects (S1 Fig) and asnap-1 was not transmitted either through the male or the female (Table 1), suggesting gametophytic lethality.

Pollen development is defective in asnap

To determine the cause for complete zero male transmission, we examined pollen development of asnap-1/+ mutants by Alexander staining for pollen cytoplasmic viability (Fig 3A–3D), by DAPI staining for the development of tricellular pollen (Fig 3E and 3F), and by scanning electron micrographs (SEMs) for pollen morphology (Fig 3G–3J). Half of the pollen grains from asnap-1/+ were aborted either by Alexander staining, by DAPI staining, or by SEM (Figs 3 and S2), indicating that functional loss of ASNAP caused pollen abortion. To determine exactly what occurred during pollen development by ASNAP loss-of-function, we performed plastic embedding and transverse sectioning (Fig 3K and 3L), as well as ultrastructure studies of anthers at different developmental stages (Fig 3M–3Q). In wild type, microspores in stage 10 anthers are unicellular, containing a large central vacuole with electron-dense materials inside (Fig 3K and 3M); microspores in stage 11 anthers contain a generative cell, a vegetative nucleus, as well as numerous small vacuoles (Fig 3K and 3O). By contrast, in asnap-1/+ anthers at stage 10, some of the unicellular microspores showed the detachment of cytoplasmic contents from pollen coat and disrupted organization of intracellular structures (Fig 3L and 3N). In asnap-1/+ anthers at stage 11, some microspores did not go through PMI (Fig 3P). Instead, they showed disintegration of internal organization (Fig 3L and 3P). In asnap-1/+ anthers at dehiscing stages, half of the microspores were degenerated (Fig 3L and 3Q). The defective pollen development in asnap-1/+ was confirmed by CLSM optical sections of developing anthers (S2 Fig). These results suggested that PMI was arrested by ASNAP loss-of-function, resulting in complete abortion of asnap-1 microspores. Consistent with the gametophytic defects, the deposition and organization of pollen coats were unaffected in asnap-1/+ (Fig 3I, 3J, 3N,3P and 3Q).

Defective embryo sacs by ASNAP loss-of-function fail to attract pollen tubes

To determine for the cause of reduced female fertility in asnap-1/+, we performed CLSM of ovules at various developmental stages. Optical sections of developing embryo sacs indicated that the formation of FM was not affected in asnap-1/+ plants (Fig 4A and 4D). In wild type, a FM goes through three rounds of mitosis and finally develops into a mature embryo sac with a central cell, an egg cell, and two synergid cells (Fig 4C). By contrast, half of the ovules in asnap-1/+ pistils had a defective embryo sac containing only one, sometimes two nuclei (Fig 4F) due to the defects of the first mitosis (Fig 4E). To confirm the embryo sac developmental defect, we introduced an egg cell-reporter transgene DD45p:GUS into asnap-1/+. In DD45p:GUS plants, all mature ovules were positive for GUS signals, indicating the presence of egg cells (Fig 4G). By contrast, only half of the ovules in DD45p:GUS;asnap-1/+ pistils were positive for GUS signals (Fig 4H), indicating the defective embryo sac development of asnap-1. Consistently, the transgene ES1p:NLS-YFP in wild type labeled all 7–8 nuclei of each embryo sac whereas labeled mostly one nucleus in embryo sacs of half ovules of asnap-1/+ (Fig 4H). These results demonstrated that functional loss of ASNAP caused the arrest of mitosis during female gametophytic development.

To determine whether the defective female gametophytic development by ASNAP loss-of-function resulted in female sterility, we pollinated emasculated asnap-1/+ pistils with LAT52p:GUS pollen, which allows histochemical GUS staining and examination of pollen tube growth and guidance in vivo. At 12 HAP, histochemical GUS staining of wild-type pistils showed that most ovules were targeted by a pollen tube (Fig 4I). By contrast, less than half asnap-1/+ ovules were targeted by a pollen tube at the same stage (Fig 4J). By examining wild-type (Fig 4K) or asnap-1/+ pistils (Fig 4L) emasculated and pollinated with wild-type pollen at 48 HAP, we confirmed that half of the asnap-1/+ ovules could not be fertilized (Fig 4L). These results demonstrated that defective embryo sac development by ASNAP loss-of-function resulted in the failure of pollination and of fertilization.

Defective gametophytic development of asnap-1 is mimicked by gametophytic downregulation of ASNAP and rescued by genomic ASNAP

To provide evidence that ASNAP is essential for gametophytic development, we generated an artificial microRNA construct for ASNAP, driven by a gametophytic linage promoter GPR1p [36]. The expression of amiR-ASNAP resulted in reduced seed set (Fig 5A–5E and 5P), defective pollen development (Fig 5F–5N and 5Q–5R), mimicking the male and female gametophytic defects of ASNAP loss-of-function. By quantitative real-time PCRs (RT-qPCRs), we confirmed that the GPR1p:amiR-ASNAP transgene did reduce the expression levels of ASNAP (Fig 5Q). These results supported an essential role of ASNAP in gametophytic development.

Because ASNAP is constitutively expressed (Fig 1), we wondered whether downregulating ASNAP in sporophytic tissues could affect plant growth. To this purpose, we generated UBQ10p:amiR-ASNAP transgenic plants. Transcript analysis verified the downregulation of ASNAP in different transgenic lines (S3 Fig). Two lines representing medium or strong downregulation of ASNAP were used for further analysis (S3 Fig). Downregulating ASNAP compromised plant growth (S3 Fig). Fertility of the UBQ10p:amiR-ASNAP transgenic plants was significantly reduced (S3 Fig). However, unlike that of asnap/+, UBQ10p:amiR-ASNAP transgenic plants produced pollen with defective pollen coat structure (S3 Fig), indicating sporophytic defects. These results support a role of ASNAP in sporophytic tissues in addition to that in gametophytes.

Because of the male and female gametophytic lethality, the T-DNA of Cas9-ASNAP had to be retained to ensure genomic editing on ASNAP at the following generation. To solve this problem and also to provide more evidence that asnap-1 was indeed a loss-of-function allele of ASNAP, we introduced a Cas9-resistant genomic sequence of ASNAP (ASNAPg) into asnap-1/+, in which the Cas9 target site was mutated without affecting the corresponding amino acids. We obtained wild-type-like plants with the ASNAPg;asnap-1 genotype from the transformants (S4 Fig), indicating a full complementation of asnap-1. The ASNAPg;asnap-1 plants expressed crASNAP at a level comparable to the endogenous ASNAP in wild type by RT-qPCRs (S4 Fig). By examining seed set (Fig 6A, 6B and 6Q) and pollen development (Fig 6F, 6J, 6N, 6R and 6S), we confirmed that the expression of crASNAP by introducing the genomic ASNAP sequence fully rescued the gametophytic lethality of asnap-1, further confirming the identity of asnap-1 as the null mutant allele of ASNAP.

Arabidopsis ASNAP encodes two functional isoforms

A close examination of the ASNAP genomic locus indicated that two splicing variants are encoded by ASNAP, both forms are constitutively expressed in various tissues and developmental stages by RT-qPCRs (S5 Fig). The second splicing form in Arabidopsis, ASNAP.2, encodes a smaller protein with an N-terminal truncation compared with ASNAP.1 (S5 Fig). Interestingly, similar N-terminal sequences were reported to mediate the interaction of yeast Sec17 or human α-SNAP with an integral membrane protein syntaxin [37], suggesting a functional distinction between ASNAP.1 and ASNAP.2.

To verify the functionality of two splicing variants, we introduced the coding sequences of crASNAP.1 or crASNAP.2 into asnap-1/+ using the constitutive promoter UBQ10p. We obtained homozygous asnap-1 plants expressing crASNAP.1 but not crASNAP.2, although both transgenes were expressed to a comparably high level (Figs 6Q, 6P and S6). Compared to wild type, the UBQ10p:GFP-crASNAP.1;asnap-1 plants were defective in root and stem growth, as well as were sterile (S6 Fig), suggesting the inability of ASNAP.1 to fully rescue the defects of asnap-1. However, asnap-1/+ plants expressing either crASNAP.1 or crASNAP.2 were obtained. Either crASNAP.1 or crASNAP.2 largely, although not fully, rescued the seed set reduction of asnap-1/+ (Fig 6C, 6D and 6Q), indicating a partial complementation of the female gametophytic development of asnap-1. The defective pollen development in asnap-1/+ was mostly rescued by either crASNAP.1 or crASNAP.2 (Fig 6G, 6H, 6K, 6L and 6O–6S), indicating a partial complementation of the male gametophytic development of asnap-1. These results suggested that both splicing variants are functional.

To determine whether the alternative splicing event was evolutionarily recurring, which would provide more support to its functional relevance, we searched other fully annotated plant genomes. The α-SNAPs in the unicellular organisms of the plant phylum, i.e. Chondrus crispus and Chlamydomonas reinhardtii, express only one isoform with all three domains comparable to yeast Sec17 (Figs 7 and S5). However, the alternative splicing of α-SNAP is detected in the genomes of different plant species, such as Physcomitrella patens, Sorghum bicolor, Zea mays, Brassica rapa, and Brassica napus (Figs 7 and S5). Interestingly, the single α-SNAP gene encoded in the human genome also produces two α-SNAP isoforms (Figs 7 and S5). These alternative splicing events produce two α-SNAPs with similar domain organizations as ASNAP.1 and ASNAP.2 in Arabidopsis, respectively (Figs 7 and S5). These results suggested that the alternative splicing of α-SNAP is an evolutionarily reoccurring event.

Subcellular localization of two ASNAP isoforms

Because Arabidopsis ASNAP.2 lacks an N-terminal sequence compared with that of Arabidopsis ASNAP.1 and yeast Sec17, we hypothesized that the two isoforms might have different subcellular localization. To test this hypothesis, examined the distribution of GFP-ASNAP.1 and GFP-ASNAP.2 by confocal laser scanning microscopy (CLSM). Root epidermal cells expressing GFP-ASNAP.1 or GFP-ASNAP.2 were pulse-labeled with the lipophilic dye FM4-64, which first indicates the plasma membrane (PM) and then is internalized to different endomembrane compartments [38]. Examination of GFP-ASNAP.1 indicated that GFP-ASNAP.1 was present in the cytoplasm as well as punctate vesicles, which partially co-localized with the internalized FM4-64 after 30 min uptake (Fig 8), indicative of the trans-Golgi network/early endosome (TGN/EE). Indeed, treatment of root epidermal cells with Brefeldin A (BFA), a fungal toxin that causes the accumulation of TGN/EE vesicles into so-called BFA compartments, resulted in the co-localization of GFP-ASNAP.1 and FM4-64 into BFA compartments (Fig 8), confirming that a portion of TGN/EE-associated GFP-ASNAP.1. To provide further evidence that a portion of ASNAP.1 associates with the TGN/EE, we introduced HAP13g:mRFP [39] into the UBQ10p:GFP-ASNAP.1-transgenic plants. CLSM of the HAP13g:mRFP;UBQ10p:GFP-ASNAP.1 plants showed that GFP-ASNAP.1 was partially colocalized with HAP13-mRFP (S7 Fig), a marker for the TGN/EE. On the other hand, wortmannin (WM) caused the formation of FM4-64-positive rings, previously reported to be enlarged prevacuolar compartments/multivesicular bodies (PVC/MVB) [40]. These ring-shaped compartments contained also GFP-ASNAP.1 (Fig 8), indicating that a portion of ASNAP.1 associated with PVC/MVB. By examining WAVE22R;UBQ10p:GFP-ASNAP.1 plants [41], we determined that GFP-ASNAP.1 also partially associated with the Golgi apparatus (S7 Fig). These results indicated that ASNAP.1 is present both in the cytoplasm and also at various endomembrane compartments, consistent with its canonical role in the disassembly of SNARE complexes.

By contrast, GFP-ASNAP.2 was mostly present in the cytoplasm (Fig 8). Partial colocalization of GFP-ASNAP.2 with FM4-64 at the PM was also detected (Fig 8). The PM-associated GFP signals were abolished by BFA treatment (Fig 8), likely because BFA treatment enhanced endocytosis and inhibited exocytosis. However, GFP-ASNAP.2 did not accumulate into BFA-compartments positive for the co-labeled FM4-64 (Fig 8), suggesting that GFP-ASNAP.2 is not associated with the TGN/EE. In addition, GFP-ASNAP.2 was also non-detectable at WM-induced ring-like structure (Fig 8), indicating that GFP-ASNAP.2 is not associated with PVC/MVB. The distinct localization of two ASNAP isoforms suggests their functional distinction.

Both ASNAP isoforms interact with Arabidopsis NSF

Despite the reports on NSF-independent function of α-SNAP in mammals, the classic role of α-SNAP is to facilitate the disassembly of SNARE complex by forming a complex with NSF [25, 42]. By sequence homology, we identified a single gene in Arabidopsis encoding NSF, At4g04910, which is constitutively expressed [43] and whose coding sequence is homologous to yeast Sec18 and human NSF. Both in yeast and in mammals, Sec18/NSF interacts with Sec17/α-SNAP through its C-terminal residues [44, 45], which are conserved in both isoforms of Arabidopsis ASNAPs (S5 Fig). To determine whether Arabidopsis ASNAP interacts with NSF, we performed bimolecular fluorescence complementation (BiFC) assays. Indeed, both ASNAP.1 and ASNAP.2 showed interactions with NSF (S8 Fig). To verify that both ASNAP isoforms interact with NSF, we performed in vitro pull-down assays, in which his-tagged NSF was able to pull-down both ASNAP isoforms (Fig 9A). To provide further evidence for their interactions, we performed fluorescence resonance energy transfer (FRET) assays that are quantitative and allow the detection of individual interacting partners in addition to the presence of their complex. Indeed, the expression of mCherry-NSF with GFP-ASNAP.1 or GFP-ASNAP.2 showed a significant higher FRET efficiency than that of mCherry with GFP-ASNAP.1 or GFP-ASNAP.2 (Fig 9B and 9C). The interaction between ASNAPs and NSF in Arabidopsis suggests an evolutionarily conserved way of function for the SNARE-disassembly complex.

Discussion

In this study, we demonstrated that Arabidopsis ASNAP is an essential gene for both male and female gametophytic development. The development of asnap microspores starts to show defects during PMI (Fig 3). At this stage, wild-type microspores undergo dynamic vacuolar re-organization such that a large central vacuole is fragmented into numerous small vacuoles [46]. Similarly, the development of asnap female gametophytes is arrested before the first mitotic division (Fig 4) when each wild-type FM produces two nuclei separated by a large central vacuole [47, 48].

Although it is still unclear whether and how vesicular dynamics affect the first mitosis during male or female gametogenesis, studies in recent years suggested a direct link between defective vacuolar dynamics and gametophytic mitosis [13–15, 23, 49, 50]. Functional loss of Arabidopsis VACUOLELESS GAMETOPHYTES (VLG) compromised vacuolar formation and fusion [49]. Its mutations resulted in defective gametophytic development at similar stages to those of asnap [49]. A few other mutants in which vacuolar trafficking was compromised also showed defective gametophytic development, such as the mutants of AP-1μ/HAPLESS13 [13], the mutants of PI(3,5)P₂-metabolizing enzymes [14, 15], as well as the mutants of COPII complexes [50, 51].

ASNAP loss-of-function could not be transmitted either through the male or the female (Table 1). The other gene in Arabidopsis whose functional loss results in the same zero male and female transmission is YKT61 [23]. Interestingly, yeast YKT6, the homolog of Arabidopsis YKT61, plays an essential role in SNARE-complex-mediated membrane fusion, antagonistic with the SNARE-disassemble complex αSNAP/NSF [24]. It was reported that mutations at SNARE-coding genes, such as SEC22 [18], BET11 and BET12 [17], VAM3/SYP22 and PEP12/SYP21 [19], as well as VAMP721 and VAMP722 [22] all compromised gametophytic development, highlighting the essential roles of fine-tuned SNARE-dynamics in ensuring plant fertility.

We demonstrated that Arabidopsis ASNAP encodes two isoforms (Figs 7 and S7). Although both isoforms interact with NSF (Figs 9 and S8), they may have distinct functions. By confocal imaging with fluorescence probes, we showed that ASNAP.1 associates with various endomembrane compartments, such as the TGN/EE, Golgi, PVC/MVB whereas ASNAP.2 is distributed mostly to the cytoplasm in addition to the PM (Figs 8 and S7). Introducing either ASNAP.1 or ASNAP.2 mostly restored male and female fertility of asnap-1/+ (Fig 6), suggesting that both isoforms are functional. In addition, ASNAP.1-transgenic plants with the homozygous asnap-1 background grew poorly (S6 Fig), indicating that both isoforms are needed for sporophytic growth. In addition, the presence of similar alternative splicing of α-SNAP in human and other plant species indicates that functional distinction of two α-SNAP isoforms is evolutionarily conserved.

Materials and methods

Plant growth and transformation

Arabidopsis Columbia-0 ecotype was used as wild type for all experiments. Mutants including asnap-1/+、asnap-2/+、asnap-3/+ were generated by CRISPR-Cas9 [35]. Plants were grown as described [52]. Stable transgenic plants were selected on half-strength MS medium supplemented with 30 μg/ml Basta salts (Sigma-Aldrich) or 25 μg/ml Hygromycin (Roche). Transgenic plants including LAT52p:GUS [53], DD45p:GUS [39], and ES1p:NLS-YFP [54, 55] were described previously.

DNA manipulation

All constructs were generated using the Gateway technology (Invitrogen) except for CRISPR/Cas9 constructs. pENTR/D/TOPO (Invitrogen) was used to generate all entry vectors. Full-length genomic sequence of ASNAP was cloned by using the primer pair ZP5533/ZP5535. Then the sequence was introduced into the destination vector GW:GUS [52] to generate the expression vector ASNAPg:GUS. The full-length CDS of ASNAP.1 or Cas9-resistant ASNAP.1 (crASNAP.1) was cloned by using the primer pair ZP10000/ZP10001 or ZP9284/ZP9285/ZP9286/ZP9287, respectively. The full-length CDS of ASNAP.2 and Cas9-resistant ASNAP.2 (crASNAP.2) was cloned by using the primer pair ZP333/ZP397 or ZP9284/ZP9285/ZP9286/ZP9287, respectively. Entry vectors were used in LR reactions with the destination vector UBQ10p:GFP-GW and 35Sp:GFP-GW [13, 56] to generate UBQ10p:GFP-ASNAP.1, UBQ10p:GFP-crASNAP.1, UBQ10p:GFP-ASNAP.2, UBQ10p:GFP-crASNAP.2, 35Sp:GFP-ASNAP.1, and 35Sp:GFP-ASNAP.2. The full-length CDS of NSF was cloned by using the primer pair ZP9294/ZP9295. Entry vector for NSF was used in LR reactions with the destination vector 35Sp:mCherry-GW to generate 35Sp:mCherry-NSF.

For the CRISPR/Cas9 construct used to generate the asnap mutants, the target site on ASNAP was selected using an online bioinformatics tool (http://www.genome.arizona.edu/crispr/CRISPRsearch.html) and was incorporated into forward and reverse PCR primers. The ASNAP-CRISPR/Cas9 cassette was generated by PCR amplifications from pCBC-DT1T2 [35] with the primers ZP5199/ZP5200/ZP5201/ZP5202. PCR products were digested with BsaI and inserted into pHSE401, resulting in pHSE401-ASNAP. To verify that the CRISPR-Cas9 construct resulted in the genomic editing of ASNAP, the primer pair ZP5203/ZP5204 were used to amplify the genomic sequences of pHSE401-ASNAP-transformed plants. The primer ZP5203 was used to sequence the amplified genomic fragment. For the amiR-ASNAP construct, the target site and sequence-specific primers for ASNAP were determined using an online tool (http://wmd3.weigelworld.org/cgi-bin/webapp.cgi). The amiR-ASNAP cassette was generated by PCR amplifications from pRS300 with the primers ZP9288/ZP9289/ZP9290/ZP9291. The resultant PCR products were cloned into pENTR/D/TOPO. The entry vector was used in LR reactions with the destination vector GPR1p:GW-GFP and UBQ10p:GW-GFP.

Constructs used in BiFC assays were generated using the destination vectors pSITE-cEYFP-C1, pSITE-nEYFP-C1, pSITE-nEYFP-N1 [57]. Expression vectors used in in vitro pull-down assays were generated by double digestions and ligations. Coding sequences were amplified with the following primer pairs: ZP10836/ZP10837 for NSF, ZP10961/ZP10962 for ASNAP.1, and P732/P733 for ASNAP.2. PCR products were digested either with BamHI/SalI (for NSF) or with BamHI/XhoI (for ASNAP.1 and ASNAP.2). Digested fragments were inserted into the destination vector pET-32a [58] pre-digested with BamHI/SalI or BamHI/XhoI using the pEASY-Uni Seamless Cloning and Assembly Kit (TRAN). Constructs were sequenced and analyzed using Vector NTI. All PCR amplifications were performed with Phusion hot-start high-fidelity DNA polymerase with the annealing temperature and extension times recommended by the manufacturer (Thermo Fisher Scientific). All primers are listed in S1 Table.

RNA extraction and RT-qPCRs

Total RNAs were extracted by using a Qiagen RNeasy plant mini kit according to the manufacturer’s instructions. Oligo (dT)-primed cDNAs were synthesized by using SuperScript III reverse transcriptase with on-column DNase digestion (Invitrogen). For RT-qPCRs of ASNAP at diverse tissues, total RNAs were isolated from seedlings and roots at 7 DAG, from leaves at 14 DAG, from stems at 25 DAG, or from reproductive tissues at 4–5 days after anthesis. For RT-qPCRs analyzing the expression of ASNAP in GPR1p:amiR-ASNAP, RNAs were extracted from inflorescences. RT-qPCRs were performed with the Bio-Rad CFX96 real-time system using SYBR Green real-time PCR master mix (Toyobo) as described [52]. Primers used for RT-qPCRs are the following: ZP9086/ZP9087 for the endogenous ASNAP, P56/P57 for ASNAP.1, P114/P115 for ASNAP.2, and ZP12/P53 for the exogenous ASNAP. Primers for GAPDH and ACTIN2 in RT-qPCRs were as described [52]. All primers are listed in S1 Table.

Biochemical assays

For the purification of recombinant proteins in in vitro pull-down assays, GST-ASNAP.1, GST-ASNAP.2, or His-NSF were transformed into E. coli strain BL21 (Rosetta), cultured at 37°C in Lurani-Bertani medium at the presence of antibiotics (100 mg/mL ampicillin) to an OD₆₀₀ of 0.6 to 0.8. Protein expression was induced by adding 0.8 mM isopropyl-b-D-1thiogalactopyranosid (IPTG). In vitro pull-down assays were performed as described [8, 39, 58]. The recombinant proteins were affinity-purified according to the manufacturer’s protocol (GE Healthcare Life Science) and analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) as described [58].

BiFC and FRET assays

BiFCs were performed in tobacco (Nicotiana tabacum) by transient transformations as described [58–60]. Constructs expressing mRFP-fused endomembrane marker proteins, including the tonoplast-associated INT1 [13] and PM-associated CBL1 [61] were described. For FRET assays, the vectors 35Sp:GFP-ASNAP.1, 35Sp:GFP-ASNAP.2, 35Sp:mCherry-NSF and 35Sp:mCherry were performed in Arabidopsis protoplasts by transient transformations as described [58, 62]. The calculation of FRET efficiency is as described [62].

Phenotypic analysis

Pollen development, including Alexander staining, DAPI staining, SEM, transverse section, and TEM of developing anthers were as described [13, 14, 52, 53, 63–65]. Histochemical GUS analysis of LAT52p:GUS-pollinated pistils and aniline blue staining of pollinated pistils were performed as described [53, 65]. Methods to analyze ovule development including whole-mount ovules clearing, optical sections of developing flowers, and examination of marker-expression in embryo sacs were as described [8, 39, 55, 59].

Fluorescence microscopy and pharmacological treatment

FM4-64 staining of root epidermal cells [13, 61, 66, 67] and Lysotracker red-staining of ovules [39, 59] were as described. Fluorescent images were captured using a Zeiss LSM 880 confocal laser scanning microscope (CLSM) with a 40/1.3 oil objective. GFP-RFP double-labeled materials were captured alternately using line-switching with the multi-track function (488-nm for GFP and 561 nm for RFP). Fluorescence was detected using a 505- to 550- nm filter for GFP or a 575- to 650-nm band-pass filter for RFP. YFP-RFP double-labeled materials were captured alternately using line-switching with the multi-track function (514 nm for YFP and 561 nm for RFP). Fluorescence was detected using a 520- to 550-nm band-pass filter for YFP or a 575- to 650-nm band-pass filter for RFP. Differential interference contrast (DIC) imaging of ovules were performed using a Zeiss Axiophot microscope with DIC optics. Image processing was performed with the Zeiss LSM image processing software (Zeiss).

Phylogenetic analysis and genomic structure

Multiple sequence alignments were performed using the MEGA7 software package and VectorNTI. An unrooted phylogenetic tree was calculated with the neighbor-joining method, and tree topology robustness was tested by bootstrap analysis of 1,000 replicates. Alignment analysis of ASNAPs were performed by using VectorNTI software. All parameters correspond to default definitions.

Statistical analysis

Quantification data are analyzed by using GraphPad Prism 6.02 (www.graphpad.com/scientific-software/prism/). All statistical analyses, One-Way ANOVA (Tukey’s multiple comparisons test) and t-test, were performed with build-in analysis tools and parameters.

Accession numbers

Arabidopsis Genome Initiative locus identifiers for the genes mentioned in this article are: At3g56190 for ASNAP and At4g04910 for NSF.

Supporting information

S1 Fig [a]
Reduced seed set of /+ is due to female gametophytic defects.

S2 Fig [a]
loss-of-function compromises pollen development.

S3 Fig [a]
Downregulating constitutively compromised plant growth and fertility.

S4 Fig [a]
Functional loss of is fully rescued by a Cas9-resistant ASNAP genomic fragment.

S5 Fig [a]
Arabidopsis encodes two isoforms.

S6 Fig [a]
plants are defective in growth and fertility.

S7 Fig [pdf]
ASNAP.1 is associated with endomembrane compartments.

S8 Fig [bifc]
Both ASNAP isoforms interact with NSF.

S1 Table [pdf]
Oligos used in this study.