Arabidopsis thaliana plants were grown in a greenhouse at 21°C and 16-h light/8-h dark photoperiod. T-DNA insertion mutant of kcs7 (At1g71160: kcs7-1, SAIL_207_H11; kcs7-2, SALK_023400), kcs15 (At3g52160, SALK_209946C), and kcs21 (At5g49070: kcs21-1, SALK_005116; kcs21-2, SALK_089611) in Col-0 background were all obtained from the ABRC stocks. T-DNA insertion status of KCS7, KCS15, and KCS21 was confirmed using the T-DNA border primer in combination with gene-specific primers listed in Supplementary Table 1.

HMMER search algorithm (E-value = 1e−10) and NCBI Basic Local Alignment Search Tool algorithm (BLASTP) (E-value = 1e−10) were used to search for KCS proteins (Finn et al., 2016). The protein sequences from A. thaliana, Capsella grandiflora, Carica papaya, Oryza sativa, Zea mays, Amborella trichopoda, Ginkgo biloba, Physcomitrella patens, and Chlamydomonas reinhardtii genomes were extracted from Phytozome v10¹ and were further verified using the SMART database.² The full-length amino acids were used to construct a phylogenetic tree with the maximum likelihood (ML) method based on the JTT amino acid substitution model.

The total RNA was isolated using TRIzol reagent (Invitrogen, Thermo Fisher Scientific, United States) according to the instructions of the manufacturer. The RNA quantity and purity were assessed using NanoQuant Spectrophotometer (Tecan, Switzerland). The first-strand cDNAs were synthesized from DNase I-treated total RNA using the cDNA Synthesis Kit (Takara, Japan). The gene expression was normalized using tubulin beta 8 (TUB8) (At5g23860) as the reference gene (Gu et al., 2014), and the relative gene expression was calculated as the mean of three biological replicates and three technical replicates. Gene-specific primers used for quantitative reverse transcription-PCR (qRT-PCR) are listed in Supplementary Table 1.

Pollen viability was assessed using Alexander’s staining. Stage 12 anthers were collected and stained with Alexander’s solution overnight at room temperature. Then, the anthers were photographed using a microscope when the active pollen grains were stained red. For scanning electron microscopic (SEM) analysis, mature pollen grains from freshly dehisced anthers were mounted and coated with gold on stubs. The samples were then immediately observed using FESEM (SU8010, Japan). Transmission electron microscopy (TEM) analysis was performed as described previously (Zhang et al., 2007). Different stage anthers were fixed in 0.1 M phosphate buffer (pH 7.2) with 2.5% glutaraldehyde (v/v) buffer and embedded into resin. Ultrathin sections (70 nm thick) were cut, counterstained with 2% uranyl acetate in 100% methanol and 1% dimethyl sulfoxide, and observed using a TEM microscope (JEOL, Japan). For pollen coat deposition analysis, more than 20 anthers were cut and photographed using TEM. The procedure of semithin sectioning and lipid staining was performed as described previously (Zhang et al., 2007; Lou et al., 2014). Semi-sections (1 μm thick) were stained using Tinapol (Sigma, United States) for 15 min (10 μg/μl) and diethyloxadicarbocyanine iodide (DiOC₂) for 5 min (5 μg/ml), and then were photographed using Olympus BX51 Microscope (Japan) using 390–440 nm excitation filter and 478 nm blocking filter.

For protein localization analysis, the full genomic sequences including the promoter of KCS genes were cloned and ligated with the p1300-GFP vector (Supplementary Table 1). The recombined plasmids were transformed into col wild-type (WT) Arabidopsis using the floral dip method (Clough and Bent, 1998), and the transgenic plants were screened using 20 mg L^–1 hygromycin. The flower buds of the T₂ transgenic plants were harvested, and the different stage anthers were isolated, and the green fluorescent protein (GFP) fluorescence was scanned using Zeiss LSM 510 confocal scanning microscope. The GFP excitation was detected at 488 nm, and the emitted light was detected at 520 nm. Ten anthers were analyzed for each stage.

For pollen hydration, freshly opened ms188 pistil was cut and attached upright on agar (1.0%). Then, pollen grains (n = 50–80) from a freshly dehiscing anther were transferred onto the stigmatic papilla. The behavior of the pollen was captured under a microscope immediately after pollinations were initiated. For in vivo pollen tube growth analysis, pollinations were initiated on ms188 stigmas and allowed to proceed for 2 h as described previously (Wang et al., 2017). Then, stigmas were fixated overnight (60% v/v ethanol, 30% v/v chloroform, and 10% v/v acetic acid), incubated in 8 M NaOH for 20 min, and washed in dH₂O three times (5 min/each). Samples were dyed in 0.1% decolorized aniline blue (0.1% w/v aniline blue in 0.1 M K₃PO₄, pH11) for 1 h, and images were captured using an Olympus BX51 microscope (Japan). Pollen tube growth was calculated as the mean of three biological replicates. All statistical analyses were carried out using Microsoft Excel.

More than 500 pollen grains were photographed using SEM, and the aborted pollen grains were numbered. The relative fluorescence fold change of DiOC₂ was quantified by measuring 80 pollen grains from 6 differential anther sections using Quantity One software (Bio-Rad, United States). The relative fold change of pollen coat concentration was quantified by measuring 60 pollen grains from 20 differential TEM sections using Quantity One software (Bio-Rad, United States). One-way ANOVA or Student’s t-test was used to evaluate the statistical significance between different genotypes. Different lowercase letters above the brackets represent statistically significant differences.

Arabidopsis genome contains 21 KCS genes. Of them, KCS1/5/6/7/9/10/11/13/15/20/21 were reported to express in flowers (Joubès et al., 2008). We performed RT-PCR analysis and confirmed their expression in flowers (Supplementary Figure 1). Among them, KCS5, KCS7, and KCS15 were specifically expressed in flowers, while KCS6, KCS9, KCS10, KCS13, KCS20, and KCS21 were relatively highly expressed in flowers (Supplementary Figure 1). To further understand the expression patterns of these KCS proteins in the developing anthers, the KCS genomic sequences were fused to the GFP gene to make the reporter constructs (pKCS:KCS-GFP) to transform WT Arabidopsis plants. Transgenic lines expressing pKCS:KCS-GFP in anthers were identified (Figure 1A). The GFP signals of KCS7-GFP, KCS15-GFP, and KCS21-GFP proteins displayed similar expression patterns (Figure 1A). Their GFP fluorescence was detected in tapetum at anther stages 8–10 (Figure 1A) and in anther locule at anther stage 10 when tapetum PCD occurs. The expression of these KCSs is quite similar to that of tapetal sporopollenin synthesis genes (Wang et al., 2018). KCS20-GFP signal was also observed in tapetum at anther stage 9 and distributed within the anther locule following tapetum PCD. KCS5-GFP signal was localized in the epidermis at anther stages 10–12, while KCS10-GFP signal was preferentially detected in both epidermis and endothecium at anther stages 10 and 11 (Figure 1B). These demonstrate the expression of four KCS members, namely, KCS7, KCS15, KCS20, and KS21, in the tapetum.

Dysfunctional tapetum 1, TDF1, aborted microspore (AMS), MS188, and MS1 are essential tapetum regulators that form a genetic pathway (Wilson et al., 2001; Sorensen et al., 2003; Zhang et al., 2006, 2007; Zhu et al., 2008, 2011). Based on the previous microarray data, KCS7, KCS15, and KS21 act downstream of these regulators (Figure 2A; Zhu et al., 2011). We collected the inflorescences of dyt1, tdf1, ams, ms188, and ms1 for quantitative real-time PCR (qPCR) analysis. The qPCR results showed that the expressions of KCS7, KCS15, and KCS21 were downregulated, while KCS20 was not affected in these mutants (Figure 2B).

Male sterility 1 is the key regulator for late tapetum development in the genetic pathway (Zhu et al., 2011). The downregulation of KCS7, KCS15, and KCS21 in ms1 mutant suggested that these three KCS genes act downstream of MS1. To further analyze the molecular regulation between MS1 and KCS7, KCS15, and KCS21, we introduced the pKCS:KCS-GFP construct into the ms1 plants and analyzed the distribution of KCS-GFP signals (Figure 1A). The GFP signals of KCS7-GFP, KCS15-GFP, and KCS21-GFP could not be detected in ms1 mutant while they were clearly observed in the tapetal cells of WT (Figure 1A). These results are consistent with the microarray and qPCR results (Figures 2A,B). MYB99 is a direct target of MS1 and its mutant displays a slight defect in the pollen development (Alves-Ferreira et al., 2007). The expressions of KCS7, KCS15, and KCS21 were not changed in the myb99 mutant, indicating that these three KCS genes are not regulated by MYB99 (Figure 2C). MS188 is an upstream regulator of MS1 (Zhu et al., 2011). Our recent study showed that many anther PCPs are downregulated in both ms188 and ms1, and the expression of MS1 driven by the MS188 promoter in ms188 mutant (ms188/pMS188:MS1) restores their expression (Lu et al., 2020). We performed qRT-PCR in the same transgenic line (ms188/pMS188:MS1). The result showed that the expressions of KCS7 and KCS15 were fully restored in this transgenic plant (Figure 2D). However, the expression of KCS21 did not show any recovery compared with that in the ms1 mutant (Figure 2D). These results show that KCS7, KCS15, and KCS21 act downstream of MS1 while the mechanism of their expression regulation may be varied.

To investigate the functions of the KCS7, KCS15, and KCS21 in the anther development, the mutants were obtained from TAIR resource³ with T-DNA being inserted in the exon of KCS7, KCS15, and KCS21, respectively (Supplementary Figure 2). No obvious vegetative or reproductive morphological abnormalities were observed in any of the single (kcs7-1, kcs7-2, kcs15, kcs21-1, and kcs21-2), double (kcs7-1/15, kcs7-2/15, kcs7-1/21-1, kcs7-2/21-2, kcs15/21-1, and kcs15/21-2), and triple mutants of KCS7, KCS15, and KCS21 (kcs7-1/15/21-1 and kcs7-2/15/21-2). Aborted pollen grains were occasionally observed in kcs15 single mutant (70 out of 1,030), kcs7-1/15 (42 out of 510), kcs7-1/21-1 (36 out of 735), and kcs15/21-1 (42 out of 473) double mutants (Figure 3 and Supplementary Figure 3). However, the statistical analyses showed that the percentage of aborted pollen grains was increased to 13% in kcs7-1/15/21-1 (81 out of 630) and kcs7-2/15/21-2 triple mutant (110 out of 853, Figure 3 and Supplementary Figure 3), indicating that KCS7, KCS15, and KCS21 play redundant roles in the pollen development.

Given that both kcs7-1/15/21-1 and kcs7-2/15/21-2 triple mutants displayed a similar defect in pollen grains, kcs7-2/15/21-2 was used for further analysis. Lipids are deposited in the pollen coat during pollen development, and KCS proteins are responsible for lipid synthesis. To analyze whether mutation in KCS7, KCS15, and KCS21 affects the pollen coat lipids, the SEM and TEM observations were used. The WT pollen grains are uniformly spheroid (13.6 μm in diameter) with a well-organized exine and a smooth pollen coat. However, in the WT pollen grains of the triple mutant, the well-smoothed pollen coat is disturbed, with many particles deposited in the cavity of pollen exine (Figure 4). In addition, approximately 13% (98 out of 760) of the triple mutant pollen grains are collapsed with a shrunken morphology (6.3 μm in diameter) and displayed disorganized exine covered with additional unknown materials (Figure 4). The DiOC₂ was used to stain the fatty acid content of the pollen wall, which exhibits a red fluorescence (Gu et al., 2014). The WT mature pollen grains are deeply stained by DiOC₂ with strong red fluorescence in exine and pollen coat (Figures 4b,c,h), whereas kcs7/kcs15/kcs21 mature pollen shows weaker pink signals on pollen walls. In addition, compared with the WT pollen, the fluorescence signal is much patchy and the relative fluorescence fold change is low in kcs7/kcs15/kcs21 pollen walls (Figures 4b,c,h). In contrast to the regular arrangement of pollen coat in the WT exine (Figures 4d,f,i), phenotypic and statistical analyses showed that the kcs7/kcs15/kcs21 mutant displayed a decreased pollen coat with a disordered arrangement (Figures 4e,g,i). Therefore, these results indicate that lipid content is reduced in the pollen coat of kcs7/kcs15/kcs21.

Pollen coat is generally considered to involve in pollen-stigma interaction. A pollen hydration assay was carried out to analyze if the reduced pollen coat in kcs7/kcs15/kcs21 affects the pollen-stigma interaction. Ms188 is a male sterile line without any pollen inside anther while its stigma is not affected (Zhang et al., 2007). The pollen hydration assay was carried out by pollinating stigmas of ms188 with pollen grains from WT and the kcs7/15/21 triple mutant (Figure 5). After landing on stigma, the WT pollen began to absorb water and became spherical (pollen hydration) in about 5 min (Figure 5A). However, the hydration of the kcs7/15/21 triple mutant pollen occurred in about 10 min after pollination. Accordingly, pollen germination of the triple mutant is also delayed (Figure 5A). To determine whether the hydration delay is derived from a defect in water absorption, an in vitro hydration of WT and kcs7/15/21 triple mutant pollen in PEG 3350 series was carried out (Supplementary Figure 4). The in vitro hydration was not significantly different between WT and kcs7/15/21 triple mutant pollen in PEG 3350 series, indicating that the absence of KCS7, KCS15, and KCS21 proteins did not impair the ability of pollen to absorb water. To test whether the subsequent steps of pollination were also affected in kcs7/15/21 triple mutant, we monitored pollen tube initiation and growth as previously described (Mayfield and Preuss, 2000; Supplementary Figure 5 and Figure 5B). About 20% of WT pollen grains showed a pollen-tube emergence within 20 min of pollination, while kcs7/15/21 triple mutant pollen-tube emergence was postponed to 30 min (Figure 5C and Supplementary Figure 5). After 2 h, WT pollen produced significantly longer tubes (470 μm) and penetrated into the style. However, pollen tubes of kcs7/15/21 triple mutant were much shorter (266 μm) (Supplementary Figure 5). Therefore, the pollen tube elongation is postponed in kcs7/15/21 triple mutant (Figures 5B,C, and Supplementary Figure 5). Despite the observed defects in pollen hydration and pollen tube elongation in kcs7/15/21 triple mutant, there was no significant difference in seed set in kcs7/15/21 triple mutant compared with WT plants.

Our recent study showed that KCS6 interacts with CER2/CER2L2 for the synthesis of pollen coat VLCFAs which is essential for pollen hydration (Zhan et al., 2018). Therefore, both KCS6 in endothecium and KCS7/15/21 in tapetum might contribute to the synthesis of pollen coat lipids. Currently, no efficient techniques are available to distinguish the difference between lipids synthesized by the tapetum and endothecium, respectively. We compared the expression of KCS and CER proteins during the anther development to understand the accumulation of the lipids in the pollen coat. In a previous study, we obtained transgenic lines of pKCS6:KCS6-GFP, pCER2:CER2-GFP, and pCER2L2:CER2L2-GFP (Zhan et al., 2018). All of the GFP transgenic plants were grown in the same pot and were cultured in the same condition. KCS6, CER2, and CER2L2 displayed a longer period expression in endothecium at anther stages 8–12 (Figure 6A). KCS7, KCS15, and KCS21 were expressed in tapetum (stages 8–10) (Figures 1, 6). They displayed a shorter period expression than KCS6, CER2, and CER2L2 expressed in endothecium (Figure 6A). These results suggested that the tapetum-derived lipids are synthesized and deposited in the pollen coat much earlier than endothecium-derived lipids. MS1 protein accumulates in tapetal cells at anther stages 7 and 8 (Yang et al., 2007), overlapping with that of KCS7, KCS15, and KCS21 (Figure 6A), further indicating that the expression of KCS7, KCS15, and KCS21 is dependent on MS1 (Figure 2).

To determine the evolutionary relationships among the KCS proteins, an unrooted ML phylogenetic tree was constructed across the known plant lineages (Figure 7). The number of copies of KCS genes varied considerably among plants, ranging from 4 in the green algae C. reinhardtii to 15 in P. patens (Bryophyta), 21 in A. thaliana (eudicot), 25 in O. sativa, and 26 in Z. mays (monocot). More KCS gene copies were identified in land plants than in algae, indicating that duplication of KCS genes likely occurred after land plants split from green algae. Based on the phylogenetic analyses, the KCS proteins can be divided into two subgroups, with group I containing AtKCS3, AtKCS12, AtKCS19, and the remaining proteins belonging to group II. In addition to Arabidopsis KCS proteins, different duplication events were identified, where 12 KCS (KCS 2, 3, 4, 5, 6, 8, 9, 12, 16, 17, 18, and 20) proteins were produced by independent duplications in Brassicaceae, four KCS (i.e., KCS7, KCS10, KCS15, and KCS21) proteins were originated from the duplication events that occurred in the ancestor of angiosperm, and two KCS (i.e., KCS13, KCS14) proteins were originated from independent duplication in Arabidopsis (Figure 7). These duplication events may reflect functional specialization of Arabidopsis KCS proteins. In addition, many monocot-specific duplication events were identified in rice and maize (Figure 7), indicating the functional specification of KCS proteins between dicots and monocots.

Pollen coat lipids are a major part of the pollen coat, which constitutes the outer layer of pollen. Previous investigation suggests that some pollen coat lipids are derived from endothecium (Zhan et al., 2018). Generally, it was considered that pollen coat lipids are mainly derived from tapetum (Hernández-Pinzón et al., 1999). In this study, the reduced pollen coat lipids in the triple mutant of tapetum expressed genes (KCS7, KCS15, and KCS21) suggest that these KCSs play a role in tapetum to provide pollen coat lipids (Figures 4, 6). KCS7, KCS15, and KCS21 were expressed in tapetum from stages 8 to 10 and were released to anther locule following tapetum PCD (Figure 1). It is likely that lipids begin to deposit outside the developing microspores before tapetum PCD. With tapetum PCD, all tapetum compounds, including lipids, deposit in the pollen coat. In kcs6 mutant, the amount of part VLCFAs (C₂₉ and C₃₀ lipids) was decreased in pollen coat (Fiebig et al., 2000), indicating that KCS6-mediated VLCFAs are deposited in pollen coat. The expression of KCS6, CER2, and CER2L2 in endothecium is much later than the expression of KCS7, KCS15, and KCS21 in tapetum (Figure 6A). These further support the previous possibility that lipids from endothecium deposit outside the lipids from tapetum (Zhan et al., 2018). KCSs catalyze the synthesis of C20 to C28 lipids, while CER2 and CER2L2 are required for the production of C28 to C34 lipids (Fiebig et al., 2000; Haslam et al., 2015). This indicates that the inner pollen coat lipid from tapetum is likely medium- and long-chain fatty acids (<28 carbon atoms), while the outer pollen coat lipid from endothecium is VLCFAs (>30 carbon atoms).

The pollen-stigma interaction includes pollen adhesion, hydration, and pollen tube growth. The pollen coat lipids may display a semisolid state and waterproof the pollen grain from its dispersal to its capture on a compatible stigma (Wheeler et al., 2001). Following the pollen-stigma interactions, a “footlike” structure is formed to enhance pollen-stigma adhesion (Elleman and Dickinson, 1990). During the “foot” formation, the lipids may create a capillary system to facilitate water transfer from the stigma cell to desiccated pollen, and pollen hydration occurs. However, it is difficult to separate clearly the processes of pollen adhesion and hydration in most reports. The VLCFAs from endothecium is predicted to locate the outer pollen coat (Zhan et al., 2018). In the mutant of kcs6 and cer2cer2l2, the pollen could not hydrate on stigma, resulting in male sterility of the plants (Fiebig et al., 2000; Haslam et al., 2015; Zhan et al., 2018). This hydration defects might result from adhesion defects. The delayed hydration in the triple mutant of kcs7/15/21 suggests that the tapetum-derived lipids might assist in establishing a water gradient between the pollen and stigma after pollen adhesion (Figure 5), and its mutation only postpone the pollen hydration (Figure 5B). We propose that the endothecium-derived lipid might function early in pollen adhesion, while tapetum-derived lipid might function later in pollen hydration. This hypothesis needs to be studied further. Hydration defects have also been reported in mutants of PCPs, such as EXL4, GRP17, and PCP-B (Wolters-Arts et al., 1998; Updegraff et al., 2009; Wang et al., 2017), indicating that lipids and proteins of pollen coat may work cooperatively to facilitate pollen hydration.

Tapetum provides nutrition for microspore development, secrets hydrolases for tetrad wall dissolution, and supplies materials for pollen wall and pollen coat formation during anther development. In Arabidopsis, the genetic regulatory pathway DYT1-TDF1-AMS-MS188-MS1 is important for tapetum development and function (Zhu et al., 2011; Gu et al., 2014; Lou et al., 2014, 2018). In this pathway, AMS directly regulates ABCG26 for sporopollenin transportation (Xu et al., 2010) and MGT5 to provide the Mg2⁺ for microspores development (Xu et al., 2015). MS188 directly regulates CYP703A2 and other sporopollenin biosynthesis genes for sporopollenin synthesis and sexine formation (Xiong et al., 2016; Wang et al., 2018). Sexine is the outer pollen wall where pollen coat (including pollen coat lipids) is deposited. MS1 is a transcription factor for late tapetum development directly regulated by MS188. It regulates the expression of multiple PCPs (Lu et al., 2020). MS1 also affects the expression of KCS7, KCS15, and KCS21 for pollen coat lipid synthesis (Figures 2, 6). This pollen coat lipid may provide a matrix for the assembly of pollen coat PCPs (Figure 6B). After tapetum PCD, the endothecium-derived lipids might be transported and deposited outside of the pollen coat (Figure 6B). Together, this study reveals that following outer pollen wall formation regulated by MS188, its directly regulating transcription factor MS1 modulates the downstream gene expression for pollen coat formation. This regulatory cascade is helpful to make sure that pollen wall and pollen coat are orderly synthesized during anther development.