KIL1 terminates fertility in maize by controlling silk senescence

Abstract

Plant flowers have a functional life span during which pollination and fertilization occur to ensure seed and fruit development. Once flower senescence is initiated, the potential to set seed or fruit is irrevocably lost. In maize, silk strands are the elongated floral stigmas that emerge from the husk-enveloped inflorescence to intercept airborne pollen. Here we show that KIRA1-LIKE1 (KIL1), an ortholog of the Arabidopsis NAC (NAM (NO APICAL MERISTEM), ATAF1/2 (Arabidopsis thaliana Activation Factor1 and 2) and CUC (CUP-SHAPED COTYLEDON 2)) transcription factor KIRA1, promotes senescence and programmed cell death (PCD) in the silk strand base, ending the window of accessibility for fertilization of the ovary. Loss of KIL1 function extends silk receptivity and thus strongly increases kernel yield following late pollination. This phenotype offers new opportunities for possibly improving yield stability in cereal crops. Moreover, despite diverging flower morphologies and the substantial evolutionary distance between Arabidopsis and maize, our data indicate remarkably similar principles in terminating floral receptivity by PCD, whose modulation offers the potential to be widely used in agriculture.

The maize NAC transcription factor KIL1 terminates the fertility of female flowers maize ears by promoting programmed cell death in senescing silk strands.

IN A NUTSHELL.

Background: Flowers are receptive to pollination during a limited window of time. This fertile period varies from mere hours in some species, like cacti, to weeks in others like some orchids. At the end of this period, nonpollinated flowers will senesce and die, thus ending the potential for flowers to produce seeds and fruits. Our research focuses on the regulation and functions of programmed cell death (PCD) in the context of plant development. PCD is a genetically encoded, actively controlled form of induced cell death that is indispensable for plant growth and reproduction, and might be involved in flower senescence.

Question: We set out to explore mechanisms underlying flower senescence in the important crop plant maize. Senescence in maize flowers had previously been described to involve tissue collapse at the base of the maize stigma (called silk strand), prompting us to ask if PCD might contribute to this process.

Findings: We analyzed changes in gene expression in senescent silk strands and discovered several induced transcription factor (TF) genes in this tissue. As TFs control the activity of other genes, TFs are prime candidates to understand the regulation of silk senescence. From several TFs, we identified a candidate we named KIL1 as a key regulator of silk senescence: KIL1 is a potent PCD activator, and loss of KIL1 function extended the fertile window in maize flowers. Interestingly, KIL1-related TFs also control senescence in the model plant Arabidopsis, suggesting that the control of flower senescence by PCD is evolutionarily conserved.

Next steps: The discovery of KIL1 is a first step toward understanding flower senescence in maize. Further exploring KIL1 functions may allow us to modulate maize fertility. Especially under short-term stress situations that are known to reduce maize seed set, an extended fertile period would contribute to stabilizing seed yield in new maize varieties.

Introduction

The effective pollination period describes the time window of floral receptivity for pollination and seed set between flower maturation and the onset of age-induced senescence of nonpollinated flowers (Williams, 1965). In contrast to postpollination floral senescence, which is triggered by the onset of seed and fruit development, age-induced senescence of unpollinated flowers occurs within a species-specific time window. The effective pollination period can last from mere hours to months, representing a significant adaptive trait in nature (Trunschke and St�cklin, 2017) and a selective trait in agriculture (Brantley et al., 2019).

Earlier studies established floral petal senescence as a model of age-dependent programmed cell death (PCD; van Doorn and Woltering, 2005). In ethylene-sensitive species, the gaseous phytohormone ethylene serves as a trigger for both pollination- and age-induced senescence, whereas alternative intrinsic factors control the same processes in ethylene-insensitive species (Rogers, 2013). Senescent petals exhibit a characteristic subcellular morphology and activate transcription factor (TF) cascades promoting PCD (Shibuya et al., 2016). However, other floral organs also undergo age-induced degeneration in the absence of timely pollination. Among them, the pollen-receptive stigma surface and the ovules are decisive for the maintenance or termination of the flower’s potential to set seeds and fruits (Gao et al., 2018).

Gene regulatory networks have been implicated in the control of floral organ senescence (Rogers, 2013). We previously identified the NAC (NAM, ATAF1/2, and CUC2) family TF KIRA1 (KIR1, also named ANAC074) as a major regulator of senescence-induced PCD in the floral stigma of Arabidopsis (Arabidopsis thaliana). When inducibly misexpressed, KIR1 caused leaf chlorosis, root growth arrest, and finally death of entire Arabidopsis seedlings (Gao et al., 2018). Transfected Nicotiana benthamiana leaves transiently expressing KIR1 also displayed a rapid onset of cell death symptoms, suggesting a conservation of KIR1-dependent gene regulatory networks between dicot species. Inducible overexpression of KIR1 specifically in the stigma caused an earlier onset of stigma senescence and a precocious loss of reproductive potential. Conversely, dominant loss-of-function KIR1 forms, and kir1 mutants combined with mutants in a known leaf-senescence regulator, ORESARA1 (ORE1; Kim et al., 2014), led to a striking extension of stigma life span coupled with a moderate prolongation of floral receptivity (Gao et al., 2018).

In flowering maize (Zea mays) plants, individual silk strands emanate from individual female flowers to emerge from husk leaves enveloping the female maize inflorescence (ear) during silking. Anthesis and pollen release from the separate male inflorescences (tassels) must overlap with ear silking to facilitate wind pollination of the exposed portions of the silk strands. Each silk strand originates from a single ovary and morphologically represents an extremely elongated floral stigma bifurcating at the tip. Pollen germination occurs on multicellular receptive papilla hairs covering almost the entire length of the silk. Pollen tubes grow through the papilla hairs into a specialized silk tissue called the transmitting tract (TT), two tracks of which run the length of the silk down to a single ovule (Heslop-Harrison and Shivanna, 1977). Following pollen germination, numerous pollen tubes grow through the TTs, but generally only a single pollen tube enters the ovule and achieves fertilization (Lausser et al., 2010). Successful fertilization is followed by abscission that severs the silk strand base from the developing kernel (the one-seeded maize fruit) (Heslop-Harrison et al., 1985). The formation of this kernel-silk abscission zone has been interpreted as a polyspermy barrier in maize, because it severs the silk vascular bundles (VBs) and TTs from the developing kernel (Heslop-Harrison et al., 1985), but also as a barrier to prevent infectious fungal hyphae from growing into ovaries (Snetselaar et al., 2001).

Senescence of unpollinated silk has been associated with a tissue collapse at the silk base close to the ovary (Bassetti and Westgate, 1993c). At a macroscopic level, the collapsing bases of unpollinated aging silk strands appear shrunken and resemble the tissue alterations associated with silk abscission following fertilization (Bassetti and Westgate, 1993b, 1993c). However, unpollinated senescent silk strands do not abscise but remain attached to the ovary. It was suggested that the senescence-induced collapse of the silk base nevertheless prevents growing pollen tubes from entering and fertilizing the ovary (Bassetti and Westgate, 1993b). Thus, silk senescence might be an important factor for the well-described decrease in kernel set observed upon late pollination (Bassetti and Westgate, 1993c; Anderson et al., 2004). Usually, the first silk strands to emerge from floret rings 6–10 in the so-called mid-base region of the ear are also the first to senesce after a few days, depending on genotype and environmental conditions (Bassetti and Westgate, 1993c). With the exception of the basal-most floret rings, which tend to be somewhat delayed, silk emergence follows a gradual acropetal pattern, such that the silk strands from the top-most rings emerge up to 4–10 days after the first emerging silk strands, depending on the genotype (Bassetti and Westgate, 1993,b, 1993,c; Anderson et al., 2004). Therefore, kernel set potential in maize is directly correlated with the number of functional nonsenescent silk strands present during the time of pollen availability. For hybrid seed production, asynchrony between anthesis in the pollen parent and silk emergence in the seed parent can lead to considerable yield losses (Anderson et al., 2004; Fuad-Hassan et al., 2008). Additionally, heat and drought stress can contribute to yield reduction, either by decreasing pollen viability, by delaying silk emergence, or by accelerating the senescence of already emerged silks (Bassetti and Westgate, 1993a, 1993b).

Although silk senescence has been previously described on a morphological basis (Bassetti and Westgate, 1993a, 1993b, 1993c; Anderson et al., 2004), its molecular regulation remains largely unknown. We show here that silk senescence is positively correlated with termination of receptivity in different maize inbred lines in both the glasshouse and in the field. We also find evidence for PCD as being responsible for tissue degeneration at the silk strand base that is correlated with the senescence-induced termination of silk function. Among a number of maize TFs whose encoding genes are upregulated in senescent silk strands, we identified the maize KIR1 ortholog KIR-LIKE1 (KIL1) as a major regulator of silk senescence. KIL1 misexpression causes cell death in maize, Arabidopsis, and N. benthamiana, and leads to a precocious termination of silk receptivity in maize. Conversely, kil1 loss-of-function mutants show a delayed silk senescence and increased kernel set at late pollination, indicative of an extension of silk receptivity.

Results

Silk senescence leads to progressive loss of kernel set during manual synchronous pollination

To determine the duration of floral receptivity in different maize inbred lines, we conducted a time course analysis of manual synchronous pollinations of maize ears at different time points, measured in days after visible emergence of the first silk strands (days after silk emergence [DASE]). We scored kernel set at the mid-base of the cob (ear rings 6–10), thus defining a narrow sampling region to control for the acropetal gradient of silk emergence and senescence in the ear. To facilitate quantification, we counted kernels on one side of the cob by means of a standardized imaging pipeline. When pollinating B104 inbred lines in the glasshouse 3 DASE, we obtained near-full kernel sets in the mid-base region (Figure�1, A and B; Supplemental Figure S1). At 7 DASE, kernel sets in the mid-base region decreased significantly (P < 0.05), and dropped further at 11 DASE. We rarely recorded developing kernels at 15 DASE. In line with the acropetal sequence of ear maturation and onset of silk senescence, positions toward the tip of the ear maintained kernel set potential for a longer time. However, almost no kernels developed along the entire ear at 15 DASE (Figure�1, A and B; Supplemental Figure S1), confirming a negative correlation between inflorescence age and the potential of ovaries to become fertilized and develop into kernels.

Figure 1.

Silk senescence is correlated with increasing loss of kernel set in maize. A, Kernel set in the mid-base region (dashed rectangle) at different pollination time points in DASE. Insets show progressively degenerating silk bases, arrowheads mark contact points to the ovary. B and C, One-sided kernel set in the mid-base region; data are median values plotted with interquartile intervals. Data are from one biological replicate with at least five cobs counted per time point; additional replicates are shown in Supplemental Figure S1. *P < 0.05; **P < 0.01; ***P < 0.001, as determined by Student’s t test. B, Glasshouse kernel set. C, Field trial kernel set (Zwijnaarde 2015). D, Percentage of degenerating nuclei detected by TEM. Data represent the percentage (mean � sd) of degenerating nuclei from at least 100 nuclei. Nuclei present in the TT were counted from microscopic images or three to four plants; five to eight silks per plant. *P < 0.05; **P < 0.01; ***P < 0.001, as determined by Student’s t test. E, Increased loss of cellular integrity occurs from 7 DASE, as indicated by intensifying Evans Blue staining. F, Fixed cross-section of silk strands in the 0.5 cm of silk adjacent to the ovary via modified Drews’ method revealing two VBs and TTs; insets show progressively degenerating TT cells (arrowheads point to representative cells), (G) TEM of cross-sections from the same region as in (F) shows regularly shaped TT nuclei at 3 DASE, progressive chromatin condensation, as indicated by irregular dark areas inside the irregularly shaped nuclei at 7 and 11 DASE, and nuclear disintegration, as indicated by less defined nuclear outlines at 15 DASE. n: nucleus. Scale bars, 2 cm in (A) and 2 mm in the insets; 200 μm in (E) and (F) with 10 μm in the insets; 1 μm in (G).

To test if a similar correlation can be reproduced under field conditions, we conducted pollination assays at two field trial sites: in Zwijnaarde, Belgium (2015 and 2016), and in Johnston, Iowa, USA (2017) (Figure�1C; Supplemental Figure S1, A–G). The kernel set from the mid-base region recorded in Zwijnaarde 2016 (Supplemental Figure S1D) was generally higher than in the same location in 2015 (Figure�1C), but similar to the data obtained in Iowa in 2017 (Supplemental Figure S1, A–G). On average, kernel set data recorded for the two inbred lines B104 and B73 were comparable (Supplemental Figure S1, A, B, E and F), whereas the inbred PH1V69 had a higher kernel set (Supplemental Figure S1, C and G). Despite these differences, kernel set in all inbreds grown in the glasshouse or in the field decreased as a function of inflorescence age at pollination.

Silk base degeneration is preceded by cellular features of PCD

We observed that the basal ovary-proximal area of silk strands shows macroscopic signs of tissue degeneration progressing with age. Indeed, while the bases of silk strands collected from rings 6–10 of nonpollinated B104 ears at 3 and 7 DASE appeared macroscopically intact and turgid, silk strand bases appeared increasingly collapsed and withered at 11 and 15 DASE (insets Figure�1A).

We reasoned that macroscopically perceptible tissue degeneration at the bases of senescent silk strands might be the result of age-induced PCD. To test this hypothesis, we monitored cellular viability and hallmarks of PCD throughout the course of silk senescence. Evans blue-stained (Smith et al., 1982) silk samples revealed the occurrence of cell death at the base of senescent silk strands (Figure�1E), similar to control samples that were physically destroyed by heat treatment (Supplemental Figure S2A). To gain more detailed insights into the cellular alterations accompanying silk senescence and cell death, we employed confocal laser scanning microscopy (CLSM) to image the autofluorescence of fixed tissue (Christensen et al., 1997) in transversal silk strand sections. To standardize the investigated region, we chose to image transverse sections at a distance of 5 mm from the silk base at 3, 7, 11, and 15 DASE (Figure�1F). In 3 DASE samples, the TT and parenchyma cells displayed a regular subcellular morphology with brightly fluorescent nuclei and intact cytoplasm surrounding a dark vacuole (inset Figure�1F). In control samples killed by heat treatment, this regular morphology was disrupted (Supplemental Figure S2B). From 7 DASE onward, cytoplasmic autofluorescence became less distinct in some TT cells, indicative of vacuolar collapse. At 11 DASE, larger silk parenchyma areas were affected by disintegration (Figure�1F), and many cells in the TT appeared completely empty, indicative of advanced cell corpse clearance (inset Figure�1F). At 15 DASE, the tissue collapse was exacerbated, and all cells in the TT appeared dead and empty (inset Figure�1F). Using transmission electron microscopy (TEM), we investigated cellular ultrastructures during silk senescence. Transverse sections made 5 mm from the silk base revealed that TT cells from 3 DASE samples contain rounded nuclei with regular morphology, well developed endoplasmic reticulum (ER) and mitochondria with a dense matrix (Figure�1G; Supplemental Figure S2, C and D). The first changes became manifest at 7 DASE as irregularly shaped nuclei with condensed electron-dense chromatin, enlarged ER and Golgi apparatus (GA), and mitochondria with lower electron density (Figure�1G; Supplemental Figure S2, E–G). At 11 DASE, chromatin condensation and nuclear fragmentation increased. TT cells displayed a dilated ER, altered mitochondrial shapes, disassembly of the GA, and increased vacuolization (Figure�2G; Supplemental Figure S2, H and I). At 15 DASE, nuclei were completely fragmented and degraded, the ER appeared fragmented and devoid of ribosomes, and mitochondria were swollen and internally disorganized (Figure�2G; Supplemental Figure S2, J–N). The TEM analysis demonstrated a progressive spreading of cell death hallmarks from 7 DASE onward (Figure�2D; Supplemental Figure S2O), preceding macroscopically visible tissue degeneration, which coincided with the decline in kernel set from 7 DASE onward, thus linking silk base degeneration and loss of kernel set.

Figure 2.

Pollen tube growth is impeded in senescent silks. A, Pollen tubes (PTs) grow through the papilla silk hairs to enter the TT (arrowheads) at 3 DASE. B, During silk senescence at 15 DASE, PTs grow within papilla silk hairs but can fail to invade the TT (white arrow). C, A PT growing through the silk after entering the TT at 3 DASE. D, PT blockage (black arrow) in the basal silk region at 7 DASE. E, GUS-expressing PTs (black arrowheads) in the ovary region and 3 DASE. F, Most ovaries at 15 DASE do not show signs of PT presence. G, Median PT density (number of pollen tubes per centimeter silk) gradually decreases during silk senescence. Results of one biological replicate with at least 15 silks counted is shown, two additional replicates are shown in Supplemental Figure S3. H, At 3 DASE on average 74.25% � 4.5 (se, n = 226) of ovaries show GUS staining, indicative of PT presence; at 7 DASE this percentage decreases to 22.6% � 6.2 (n = 231), to 11.7% � 1.9 (n = 210) at 11 DASE; and to 11.8% � 3.8 (n = 223) at 15 DASE. Approximately 40–55 ovaries were extracted from the mid-base region of five ears. ***P < 0.001; ns, nonsignificant, as determined by Student’s t test. The box and whisker plots display median values with interquartile intervals. Similar results were obtained from two additional replicates of the experiment, see Supplemental Figure S3. Scale bars, 100 μm.

Pollen tube growth is impeded in senescent silk strands

Next, we investigated if tissue degeneration at the silk base blocks pollen tube growth in senescent silk strands, as previously hypothesized (Bassetti and Westgate, 1993b). In a first assay, we determined the ability of pollen to germinate on young versus senescent silk strands. To this end, we pollinated the exposed apical regions of silk strands at 3, 7, 11, and 15 DASE with pollen expressing a Act1-Dpro:β-glucuronidase (GUS) reporter construct consisting of the GUS reporter gene driven by the Actin1-D promoter to visualize pollen grains and tubes (Brettschneider et al., 1997). To assess pollen germination, we collected the apical regions of silk strands 90 min after pollination and conducted GUS staining to detect pollen grains and germinated pollen tubes. Although pollen grains were able to germinate at all four time points (in DASE), pollen germination rates gradually decreased over time (Figure�2, A, B and G). Pollen tubes germinating at 11 and 15 DASE failed to invade the TT (Figure�2B) or appeared to be arrested inside the silk (Figure�2D). To quantify the number of pollen tubes reaching the ovary during silk senescence, we pollinated silk strands at 3, 7, 11, and 15 DASE with pollen from the same reporter line and inspected mid-base region ovaries by GUS staining 24 h after pollination (Figure�2, E and F). We detected GUS staining in ∼80% of ovaries at 3 DASE, whereas only 25% showed presence of pollen tubes at 7 DASE, which further decreased to only 10% at 11 and 15 DASE (Figure�2H). Our analyses show that pollen grain germination (Figure�2G; Supplemental Figure S3) is less affected by silk senescence compared to the growth of pollen tubes toward the ovary (Figure�2H), supporting the notion that tissue degeneration at the silk base contributes to the termination of the effective pollination period in maize (Bassetti and Westgate, 1993c).

Transcriptome analysis of maize silk strands reveals effective activation of genes involved in silk senescence

The importance of transcriptional regulation for the initiation of senescence and developmental PCD (dPCD) in plants has been demonstrated in several systems (Cubria-Radio and Nowack, 2019). To identify candidate transcriptional regulators of silk senescence, we employed transcriptome deep sequencing (RNA-seq) of silk bases at four progressive stages of silk senescence (3, 7, 11, and 15 DASE; Supplemental Figure S4A). Sequencing on an Illumina NextSeq instrument generated 370 million reads, of which 85% successfully mapped to version 3 of the maize B73 reference genome, resulting in close to 29,000 expressed genes (at least 1 count per million in at least three replicates; Supplemental Table S1). To assess global gene expression changes, we followed a multidimensional scaling approach that showed a clear separation of the four senescence time points (Supplemental Figure S4B). In pairwise comparisons between the 3 DASE time point and the three other senescence time points, we detected a total of 8,437 differentially expressed genes (DEGs; Log₂[fold-change (FC)] > 1 and false discovery rate [FDR] < 0.05), of which 3,764 DEGs were upregulated and 4,673 were downregulated during silk senescence (Supplemental Data Set 1). While the number of DEGs increased with senescence stage, many DEGs were consistently upregulated or downregulated during the later stages of silk senescence (Supplemental Figure S4, C and D).

To assess the biological functions of these DEGs, we performed a gene ontology (GO) enrichment analysis using the DEGs resulting from the pairwise comparison between 15 and 3 DASE (Supplemental Table S2). The most enriched GO terms among downregulated DEGs were related to photosynthesis and chlorophyll biosynthesis, results that were reminiscent of leaf senescence in maize (Sekhon et al., 2019), and to auxin transport and homeostasis, possibly pointing to a decrease in auxin signaling during senescence (Mueller-Roeber and Balazadeh, 2014). The most enriched GO term for upregulated DEGs was related to cell wall catabolism, with other enriched terms being related to defense and carbohydrate transport, again drawing parallels between silk and leaf senescence (Norholm et al., 2006; van der Graaff et al., 2006; Zhang et al., 2014).

As we identified cellular and ultrastructural features of PCD in senescing silk bases, we investigated whether silk senescence was accompanied by the upregulation of genes related to those implicated in dPCD in Arabidopsis (Olvera-Carrillo et al., 2015). After a phylogenetic analysis in PLAZA (https://bioinformatics.psb.ugent.be/plaza/versions/plaza_v4_5_monocots/), we identified one or two close maize orthologs each for several Arabidopsis dPCD-associated genes that were also strongly upregulated during silk senescence (Table�1; Supplemental Table S3). A k-means clustering analysis revealed that these genes have highly similar expression profiles during silk senescence, belonging to either cluster 5 or 6 out of the 12 defined clusters. Cluster 5 contained genes that are upregulated only late in silk senescence, while the genes in cluster 6 were already upregulated during earlier senescence stages (Supplemental Figure S4E; Supplemental Table S3). This observation relates canonical dPCD events in Arabidopsis with the degenerative processes occurring during silk senescence, suggesting the existence of a dPCD-related process occurring during silk senescence.

Table 1.

Closest orthologs of Arabidopsis dPCD genes showing a significant increase in expression during silk senescence

Name	Gene ID_V3	Gene ID_V5	Arabidopsis ortholog	15 DASE versus 3 DASEa	KIL1-OE versus WTa	KIL1-SRDX versus WTa
ZmMC9	GRMZM2G022799	Zm00001eb196150	At5g04200 (MC9)	3.29	2.53	−6.39
ZmRNS3a	GRMZM2G161274	Zm00001eb041390	At1g26820 (RNS3)	3.44	2.79	−5.59
ZmRNS3b	GRMZM2G141322	Zm00001eb328300	At1g26820 (RNS3)	7.23	4.06	−5.47
ZmDMP4a	GRMZM5G812126	Zm00001eb383510	At4g18425 (DMP4)	4.57	4.38	−7.28
ZmDMP4b	GRMZM2G070013	Zm00001eb343640	At4g18425 (DMP4)	3.91	1.79	−6.05
ZmEXI1	GRMZM2G167584	Zm00001eb027840	At2g14095 (EXI1)	4.04	3.26	−3.91
ZmSCPL48	GRMZM2G020146	Zm00001eb229790	At3g45010 (SCPL48)	3.06	1.55	−3.61
ZmBFN1	GRMZM2G168744	Zm00001eb043930	At1g11190 (BFN1)	1.66	2.90	−4.69
ZmCEP1	GRMZM2G456217	Zm00001eb039650	At5g50260 (CEP1)	8.88	3.53	−7.29

For each gene, the Log₂(FC) between the different stages of silk senescence and earliest stage of silk senescence (3 DASE) is indicated, as well as a comparison between the silk strands of transgenic lines KIL1-OE and KIL1-SRDX (see results further down) and the WT.

^aLog₂(FC) comparing condition 1 versus condition 2.

Identification of candidate transcriptional regulators involved in silk PCD regulation

Given the relevance of transcriptional regulation during senescence, we next aimed to identify candidate TFs promoting silk senescence. We selected candidate genes with increasing expression levels throughout senescence as well as high expression levels at later senescence stages (more than 100 read counts per million at 15 DASE). In total, 54 genes encoding TFs fit these criteria, among them several TF genes from the NAC family (Supplemental Data Set 2). When sorting this list by increasing Log₂(FC) between 14 DASE and 3 DASE, many NAC TF genes, including three orthologs of the Arabidopsis stigma-senescence regulator gene KIR1, appeared among the top upregulated TF genes. This discovery indicated a potential functional conservation of KIR1-like TFs in terminating the effective pollination period across large evolutionary distances and widely divergent flower morphologies between the dicot Arabidopsis and the monocot maize. We decided to focus on the role of senescence-regulated NAC TFs, including most KIR1 orthologs, which we named KIL1 to KIL6 (Figure�3C; Table�2; Supplemental Data Set 3). As KIL6 was differentially expressed during silk senescence, we omitted this gene in the following analyses.

Table 2.

Selected candidate TF genes

Gene Name	Gene_ID_V3	Gene_ID_V5	3 DASEa	7 DASEa	11 DASEa	15 DASEa	7 versus 3 DASEb	11 versus 3 DASEb	15 versus 3 DASEb
ZmNAC36 (KIL1)	GRMZM2G081930	Zm00001eb077580	3.75	10.50	74.75	473.50	1.49	4.21	6.91
ZmNAC17 (KIL3)	GRMZM2G062009	Zm00001eb185110	2.50	6.00	26.00	101.50	1.25	3.25	5.25
ZmNAC2	GRMZM2G181605	Zm00001eb330910	17.00	33.25	96.00	531.75	1.01	2.43	4.93
ZmNAC25	GRMZM2G127379	Zm00001eb405590	53.50	245.50	765.75	1249.50	2.27	3.78	4.53
ZmNAC44	GRMZM2G011598	Zm00001eb015630	14.75	55.50	152.50	496.50	1.96	3.31	5.04
ZmNAC49	GRMZM2G347043	Zm00001eb062170	1,293.25	2,598.25	3,298.25	3,783.25	1.07	1.31	1.55
ZmNAC76	GRMZM2G316840	Zm00001eb082430	52.00	194.75	456.50	765.25	1.98	3.07	3.86
ZmNAC9	GRMZM2G134073	Zm00001eb363940	340.25	787.75	895.25	1,468.50	1.20	1.32	2.08
ZmNAC51 (KIL4)	GRMZM2G140901	Zm00001eb191780	1.50	7.25	7.75	41.25	2.18	2.21	4.64
ZmNAC104 (KIL5)	GRMZM5G857701	Zm00001eb258400	481.50	569.25	514.00	583.75	0.26	0.03	0.26
ZmNAC65 (KIL2)	GRMZM2G043813	Zm00001eb426640	11.75	14.75	61.50	370.25	0.38	2.32	4.94

Differential gene expression was calculated based on pairwise comparisons between the four stages of silk senescence with thresholds of Log₂(FC) ≥ 1 and FDR < 0.05.

^aMean expression in counts per million reads.

^bLog₂(FC) comparing condition 1 versus condition 2.

KILs are sufficient to induce cell death in maize protoplasts and N. benthamiana leaves

To allow a first and rapid assessment of the effect caused by the overexpression of candidate TF genes on PCD in maize, we developed a flow cytometry-based protoplast viability assay using transfected maize mesophyll protoplasts. In this assay, the number of viable cells co-transfected with a green fluorescent protein (GFP) viability marker and a cell death-promoting TF gene will decrease relative to cells from a control population. At 4 days after transfection (DAT), the number of GFP-positive cells from a GFP/GUS-negative control population did not decrease significantly, remaining around 80%–90% of the number of viable cells at 1 DAT (Figure�3A). In contrast, the ratio of GFP-positive protoplasts for a GFP/KIR1-positive control dropped to ∼50% of the control population (Figure�3A), indicating a cell death-promoting effect of Arabidopsis KIR1 on maize protoplasts. Of the 11 maize candidate NAC TF genes tested, 8 showed a significant effect on protoplast viability, including all five KIR1 orthologs.

Figure 3.

Screen for putative dPCD-associated genes in maize that induce cell death in N. benthamiana and maize mesophyll protoplasts. A, Protoplast viability screen showing stable GFP signal in the control cell population, which was co-transfected with 35Spro:nlsGFP + 35Spro:GUS. Over 4 days, KIRA1 and eight candidates display a similar effect on protoplast viability. Data are shown as means � sd of three different experiments. P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant, as determined by Student’s t test. B, Nicotiana benthamiana leaves infiltrated with KIRA1 and eight candidate TF genes selected from the protoplast viability screen. C, Phylogenetic tree of the NAC74 family showing KIR1 from Arabidopsis and its closest maize homologs. Bootstrap values supporting each branch are indicated. Asterisks, DEGs; crosses, genes with killing activity in the protoplast assay. D, KIR1 and its maize ortholog KIL1 transactivate the AtRNS3 promoter in maize mesophyll protoplasts. Data are shown as means � SEM of three different experiments. E, Estradiol induction of KIL1 and KIR1 overexpression lines triggers root growth arrest in 5-day-old Arabidopsis seedlings (white arrowheads) and systemic cell death of the seedlings. See also Supplemental Figure S5. Scale bars, 2 cm in (B), 1 cm in (E).

Next, we tested if overexpression of the most effective candidates revealed in the protoplast system above was sufficient to induce cell death in a heterologous dicot system via N. benthamiana leaf transfection (Figure�3B). Accordingly, we infiltrated N. benthamiana leaves with constructs harboring eight selected candidate TF genes under the control of the cauliflower mosaic virus (CaMV) 35S promoter. As previously reported (Gao et al., 2018), the overexpression of Arabidopsis KIR1 induced cell death lesions 6 days after infiltration (Figure�3B). We observed a similar, and even faster, appearance of lesions in N. benthamiana leaves infiltrated with most of the KIL paralogs, with only KIL4 showing somewhat weaker effects. These data suggest that KILs can induce cell death in monocot as well as dicot species. In contrast, we failed to observe cell death lesions after infiltration with some of the other candidate genes, including ZmNAC44, ZmNAC9, and ZmNAC76 (Figure�3B).

To gain insights into the mode of action of maize KILs, we performed a series of transient expression assays designed to quantify the transactivation capacity of individual TFs on candidate promoter fragments (Vanden Bossche et al., 2013) in B73 maize mesophyll protoplasts. We investigated the induction of the Arabidopsis RIBONUCLEASE3 (AtRNS3) promoter, which was previously shown to be very effectively transactivated by Arabidopsis KIR1 in tobacco (Nicotiana tabacum) BY2-derived protoplasts (Gao et al., 2018). Notably, we detected the strongest transactivation of the AtRNS3 promoter when overexpressing KIL1, a moderate transactivation upon overexpression of KIL2, KIL3, and KIL4, and no significant activation by KIL5, despite its pro-PCD capacity revealed in the protoplast and leaf infiltration assays above (Figure�3D). To investigate the consequences of KIL1 expression in Arabidopsis, we generated estradiol-inducible KIL1 misexpression lines (H3.3pro:XVE≫KIL1-GFP, with the promoter from HISTONE H3.3 [HTR5]). Similarly to the previously published RPS5Apro:XVE≫KIR1-GFP lines driven by the RIBOSOMAL PROTEIN S5 promoter (Gao et al., 2018), estradiol induction in two independent transgenic lines first triggered root growth arrest and later widespread cell death followed by the death of entire Arabidopsis seedlings (Figure�3E; Supplemental Figure S5). These results suggest that maize KIL1 is a functional ortholog of Arabidopsis KIR1 and might function as an inducer of PCD during silk senescence in maize.

KIL1 is a key regulator of silk senescence

To test the above hypothesis, we generated maize lines with altered KIL1 function. To obtain a promoter for silk-specific gene expression, we exploited our RNA-seq data and identified GRMZM2G145461 (Zm00001eb272090, encoding an acidic class I chitinase; Hawkins et al., 2015), with silk-specific expression that increases during silk senescence. We named this gene SILK1. Notably, expression of SILK1 was considerably higher compared to that of KIL1 throughout silk senescence (Supplemental Figure S6, A and B) and was relatively low in other tissues (Supplemental Data Set 3), rendering it suitable to target silk senescence by tissue-specific overexpression approaches. We cloned the putative promoter fragment of SILK1 (SILK1pro) and used it to drive the expression of the KIL1 coding sequence to generate a silk-preferred overexpression construct (SILK1pro:KIL1, hereafter KIL1-OE). Conversely, we fused the repressive Ethylene-responsive element binding factor-associated amphiphilic repression SRDX domain (Hiratsu et al., 2003) to the C-terminus of KIL1 to generate a dominant-negative KIL1 construct (SILK1pro:KIL1-SRDX, hereafter KIL1-SRDX). Such dominant-negative TF versions have proven effective in the analysis of redundant NAC TFs in Arabidopsis by overcoming genetic redundancy (Gao et al., 2018), and they have also been shown to work in rice (Yoshida et al., 2013). To our knowledge the use of a repressive SRDX domain has not been established so far in maize, prompting us to test several SRDX motifs in a transient expression assay before adopting the modified Arabidopsis motif LDLDLELRLGFA (Hiratsu et al., 2003) that optimally repressed a known Opaque2 target promoter in maize protoplasts (Supplemental Figure S6C). We then generated several independent stable transformants with the KIL1-OE and KIL1-SRDX expression constructs in the maize B104 inbred line with high expression levels for their relevant construct (Supplemental Figure S8B). In addition, we used clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated kinase 9 (Cas9)-mediated genome editing to generate a putative null kil1 mutant with a single nucleotide insertion. This insertion caused a frameshift leading to an altered open reading frame after amino acid 12 before reaching a premature stop codon (Supplemental Figure S6, D–F).

At 3 DASE, KIL1-SRDX and kil1 mutant plants had silk phenotypes similar to the wild-type (WT) B104, apart from slightly larger silk base diameters (Figure�4E). However, the silk bases showed no sign of tissue degeneration in either genotype at 11 DASE (Figure�4A). In contrast, silk strands of KIL1-OE lines appeared visibly shorter and thinner in diameter relative to the WT (Figure�4E), and silk strands of the mid-base region degenerated already at 3 DASE (inset in Figure�4A). Moreover, KIL1-OE lines developed shorter cobs (Supplemental Figure S8A). Conceivably, expression of cell-death inducing KIL1 in pericarp and kernel tissues conveyed by the SILK1 promoter (Supplemental Figure S6A) may cause this effect. Using Evans blue staining, we observed stronger staining intensity at the silk bases of B104 WT plants from 7 DASE onward, with prominent tissue degeneration at 11 and 15 DASE (Figures�1, E and 4, B). Silk bases of the KIL1-OE lines were already completely degenerated at 7 DASE (Figure�4B), indicating that early silk senescence is caused by KIL1 overexpression. In contrast, KIL1-SRDX lines and the kil1 mutant displayed delayed silk senescence phenotypes, as indicated by delayed accumulation of Evans Blue and tissue collapse (Figure�4B). CLSM analysis at 11 DASE revealed hallmarks of progressed senescence and cell death in silk strands overexpressing KIL1. We observed more cells lacking their cytoplasm in the TT and deterioration of the surrounding parenchyma cells when comparing KIL-OE lines with control lines (Figure�4C). However, in the kil1 mutant, and more markedly in KIL1-SRDX lines, we found a delay of senescence morphologies, including maintenance of parenchyma structure and clearly visible cytoplasm in the cells of the TT at 11 DASE (Figure�4C). These results suggest that silk-specific KIL1 overexpression leads to early silk senescence morphologies and tissue degeneration, while both silk-specific KIL1-SRDX expression, as well as the recessive kil1 mutation, cause a delay of silk base degeneration.

Figure 4.

Floral receptivity and cell death in the silk base is controlled by KIL1. A, KIL1-SRDX lines and the kil1 mutant display WT cobs with thick silk strands while KIL1-OE lines have small cobs with short and thin silks (corresponding graph shown in Supplemental Figure S8). Insets show silk bases at 11 DASE. Arrowheads point to the basal part of the silk adjacent to the ovary. B, Evans Blue staining over a silk senescence time course from 3 to 15 DASE, indicating an earlier occurrence of cellular disintegration in KIL1-OE silks, and a delay of disintegration in kil1 and KIL1-SRDX. Arrowheads point to the basal part of the silk adjacent to the ovary. C, Fixed cross-section of silk strands in the basal-most 0.5 cm of the silk visualized via modified Drews’ method shows an accelerated cellular degeneration in KIL1-OE, and a delay of degeneration in KIL1-SRDX lines and the kil1 mutant. D, KIL1-upregulated genes, as shown by the Venn diagram on the left (genes upregulated in KIL1-OE and downregulated in KIL1-SRDX, compared to the WT at 11 DASE; Log₂FC > 2 and FDR < 0.05). The three Venn diagrams on the right show the overlap between KIL1-upregulated genes and genes upregulated (Log₂FC > 2 and FDR < 0.05) at the different time points of silk senescence. The representation factor (number of overlapping genes divided by the expected number of randomly overlapping genes drawn from two independent groups) and the P-value are indicated. E, KIL1-OE lines display thinner silks. The average of silk diameter is 0.69 cm � 0.01 (se, n = 15) in B104; significantly thinner in both KIL1-OE lines: 0.47 cm � 0.03 (n = 10) and 0.59 cm � 0.02 (n = 10) in KIL1-OE A and KIL1-OE B, respectively. Silk diameter in the kil1 mutant is similar to WT: 0.7 cm � 0.02 (n = 10); both KIL1-SRDX lines display thicker silks: 0.95 cm � 0.03 (n = 15) in KIL1-SRDX C; 0.88 cm � 0.02 (n = 13) in KIL1-SRDX L. *P < 0.05; **P < 0.01; ***P < 0.001), as determined by Student’s t test. Graph represents one experiment, three independent experiments showing the same trend were performed. F, Dual-luciferase assay for monitoring transient expression of putative dPCD promoters in maize leaf protoplasts. The promoter activity measured in presence of either KIL1 or GUS was normalized by the mean value of the negative control (GUS) and the resulting FC is indicated. *P < 0.05; **P < 0.01; ***P < 0.001), as determined by Student’s t test, n = 5. The graph represents the results of one biological replicate, two additional replicates are shown in Supplemental Figure S8. Note that the different scales on the y-axes are due to technical variability. Scale bars, 2 cm in (A) with 1 mm in the insets; 200 μm in (B); 100 μm in (C) with 10 μm in the insets.

KIL1 promotes PCD-associated gene expression

To assess how KIL1-OE and KIL1-SRDX transgenes affect the gene regulatory network expressed during silk senescence, we expanded our RNA-seq analysis to the silk bases of KIL1-OE, KIL1-SRDX lines and the WT at 11 DASE. We used the same experimental setup as for the silk senescence time course (Supplemental Figure S4A). We obtained a total of 327 million reads for all replicates, of which 88% successfully mapped to version V3 of the B73 maize reference genome (Supplemental Table S4). We detected around 28,000 expressed genes in all samples. A DEG analysis performed with the same parameters as the silk-senescence time course identified 8,435 DEGs between KIL1-OE and WT and 6,991 DEGs between KIL1-SRDX and WT, indicating that silk gene expression networks are substantially affected in both transgenic lines (Supplemental Data Set 4).

Next, we compared the RNA-seq datasets for KIL1-OE and KIL1-SRDX with the dataset for silk senescence. On a multidimensional scaling plot, individual replicates of KIL1-OE, KIL1-SRDX, and WT samples clustered closely together, indicating both the reproducibility of the experiment and a strong divergence in the transcriptome of the three genotypes (Supplemental Figure S7A). Relative to the WT, the KIL1-SRDX samples showed a shift to the right along the leading dimension 1, thus in the same direction as the 3 and 7 DASE samples of the silk senescence time course. This result suggested that the transcriptome of KIL1-SRDX silk strands at 11 DASE is in a less advanced stage of senescence compared to the WT at the same time point. Consistent with the phenotypic aspect of the silk (Figure�4, A–C), these data would suggest that KIL1-SRDX expression delays the expression of silk senescence-related gene regulatory networks. In contrast, the KIL1-OE samples formed a cluster that is shifted far to the left of the WT along leading dimension 1. As the silk senescence time course also progressed from right to left in this dimension, this position may indicate a more advanced stage of senescence in KIL1-OE silk strands (Supplemental Figure S7A). We obtained a similar trend by a hierarchical clustering analysis: The KIL1-SRDX samples clustered together with the 3 and 7 DASE samples, while WT samples clustered with the 11 and 15 DASE samples (Supplemental Figure S7, B and C). This observation confirmed the interpretation that KIL1-SRDX silk strands at 11 DASE are in a less advanced stage of senescence. KIL1-OE samples, however, formed an independent cluster that might correspond to a more advanced stage of senescence beyond 15 DASE (Supplemental Figure S7, B and C). Moreover, the 10 most-enriched GO terms among upregulated DEGs during silk senescence were also enriched in DEGs upregulated in KIL1-OE, and downregulated in KIL1-SRDX. Conversely, the 10 most-enriched GO terms in the DEGs downregulated during silk senescence were also enriched in both DEGs downregulated in KIL1-OE and upregulated in KIL1-SRDX (Supplemental Table S5). Finally, we identified 284 KIL1-induced genes by intersecting genes upregulated in KIL1-OE and downregulated in KIL1-SRDX (Log₂[FC] > 2, FDR < 0.05; Figure�4D). Most of these genes were also upregulated during the late stages of silk senescence in the WT (Log₂[FC] > 2, FDR < 0.05; Figure�4D). Accordingly, we found that most orthologs of Arabidopsis dPCD-associated genes are not only upregulated during silk senescence (Table�1), but were also upregulated in both KIL1-OE silk and downregulated in KIL1-SRDX silk (Supplemental Table S6). Altogether, these data suggest that KIL1 regulates a gene network promoting silk senescence, in particular senescence-induced dPCD by transcriptional activation of downstream genes.

To experimentally test this hypothesis, we determined if KIL1 is sufficient to activate the promoters of selected dPCD-associated gene orthologs in maize. We performed transient expression assays to test the promoters of maize orthologs of dPCD-associated genes in maize mesophyll protoplasts. We established that the promoters of ZmRNS3a, ZmRNS3b, and METACASPASE9 (ZmMC9) are slightly but significantly activated by KIL1, whereas the promoters of CYSTEINE ENDOPEPTIDASE 1 (ZmCEP1), BIFUNCTIONAL NUCLEASE1 (ZmBFN1), EXITUS1 (ZmEXI1), DUF679 MEMBRANE PROTEIN 4a (ZmDMP4a), and ZmDMP4b were strongly activated by KIL1 (Figure�4E). This experiment indicates that KIL1 can directly or indirectly activate the expression of orthologs of Arabidopsis dPCD-associated genes in maize, confirming KIL1 as a key regulator of senescence and senescence-induced dPCD in maize silk.

KIL1 modulation is sufficient to extend or shorten the effective pollination period

Finally, we investigated whether KIL1 modulation was not only sufficient to accelerate or delay silk senescence, but if it might also shorten or extend the effective pollination period for potential use in agriculture. To this end, we performed pollination assays as described above (Figure�1) and compared the kernel set in the mid-base region of KIL1-OE and KIL1 loss-of-function (KIL1-SRDX and kil1) plants with that of WT at 3, 7, and 11 DASE (Figure�5, A–D).

Figure 5.

Overexpression or inactivation of KIL1 reduces or increases kernel set after late pollination, respectively. A, Manual synchronous pollinations were performed at 3, 7, and 11 DASE, respectively, and kernel set was evaluated in the mid-base region of the ear (dashed rectangle). B, Kernel set in the mid-base region under greenhouse conditions at 3 DASE. Data represent median values plotted with interquartile intervals. One-sided kernel set recorded was on average 29.6 � 2.3 (sd, n = 5) kernels in the WT; 5.0 � 2.4 (n = 5) kernels in KIL1-OE; 31.4 � 1.9 (n = 5) kernels in kil1; and 32.4 � 0.7 (n = 5) kernels in KIL1-SRDX. C, Kernel set in the mid-base region under greenhouse conditions at 7 DASE. Data represent median values plotted with interquartile intervals. One-sided kernel set recorded was on average 24.8 � 2.1 (sd, n = 5) kernels in the WT; 0.2 � 0.2 (n = 5) kernels in KIL1-OE; 31.2 � 1.15 (n = 5) kernels in kil1; and 22 � 1.75 (n = 5) kernels in KIL1-SRDX. The graph shows the results of one biological replicate, two additional replicates are shown in Supplemental Figure S9. D, Kernel set in the mid-base region at 11 DASE. Data represent median values plotted with interquartile intervals. One-sided kernel set recorded was on average 0.6 � 2.1 (sd, n = 5) in the WT; 0.0 (sd = 0.0, n = 5) in KIL1-OE; 12.8 � 2.2 (n = 5) in kil1; and 15.8 � 5.7 (n = 5) in KIL1-SRDX. The graph shows the results of one biological replicate, one additional replicate is shown in Supplemental Figure S9. Scale bars, 2 cm.

We determined that kernel set in KIL1-OE plants is drastically lower already at 3 DASE, and practically absent at 7 and 11 DASE, compared to the WT (Figure�5, A–D). Conversely, both the kil1 mutant and the KIL1-SRDX lines showed no significant change in kernel set at 7 DASE; on the contrary, the kil1 mutant even showed a significant increase in kernel set at 7 DASE compared to the WT (Figure�5, A–C,; Supplemental Figure S9). At 11 DASE, both the kil1 mutant and the KIL1-SRDX lines showed a significant increase in kernel set compared to the WT. Kernel set was rather variable between biological replicates of kil1 and especially KIL1-SRDX, but some KIL1-SRDX ears showed a near-full kernel set at 11 DASE, which we never observed in the WT (Figure�5, A–D; Supplemental Figure S9).

Discussion

This study identifies KIL1 as a key regulator of silk senescence and senescence-induced dPCD in maize. Modulation of KIL1 function is not only sufficient to accelerate or delay the senescence process in unpollinated silks but also to control the ear’s effective pollination period for kernel set. Overexpression of KIL1 in silk tissue is sufficient to interfere with kernel set even during early stages of inflorescence senescence, while loss of KIL1 function delays silk senescence and extends the effective pollination period, increasing kernel set at late pollination time points.

All plant species have a genetically determined effective pollination period that underlies natural selection and is optimized with regards to several factors, including timing and abundance of pollen and pollinators, environmental conditions, as well as the presence of herbivores and pathogens (Williams, 1965; Trunschke and St�cklin, 2017; Brantley et al., 2019). An optimal effective pollination period maximizes the chances for full seed set and fruit development, while minimizing energy expenditure and water loss by maintaining floral organs, as well as lowering the chance for herbivores or pathogens to ingest or infect reproductive tissues. In agriculture, human selection to maximize seed or fruit number and size has not necessarily involved modulation of effective pollination period, offering a chance to exploit this trait in crops to maximize the chance for seed or fruit set, especially under adverse environmental conditions (Habben and Schussler, 2017).

Our results show that both KIL1 and its Arabidopsis ortholog KIR1 promote PCD to terminate the effective pollination period. Considering the large evolutionary distance between maize and Arabidopsis, and their fundamentally different flower morphologies, it is intriguing that KIR1 and KIL1 not only possess a conserved molecular role as dPCD activators, but also a similar role in controlling floral organ senescence to regulate the effective pollination period. In Arabidopsis, it is the papilla cells, which are specialized elongated epidermal cells of the floral stigma, that undergo KIR1-promoted cell death to terminate floral receptivity (Gao et al., 2018). In contrast in maize, KIL1 promotes dPCD at the base of the elongated stigma, the silk strand, to achieve the same goal. Possibly, the function of KIR1-like genes may have arisen before the split of monocots from the dicot lineage 150 million years ago and has thus remained conserved in both lineages. Alternatively, KIR1-like genes might have been independently recruited to promote cell death in different parts of the flower to fulfill an analogous function.

Interestingly, a KIL1 ortholog was identified as the long-hypothesized D gene in sorghum (Sorghum bicolor), where the encoded protein acts as a master transcriptional regulator to induce programmed death of stem pith parenchyma cells (Fujimoto et al., 2018). Furthermore, KIL1/NACTF36 was recently found among the upregulated genes associated with PCD-mediated carpel suppression in tassels (Klein et al., 2022). These findings suggest that KIL1 orthologs are key regulators of developmentally controlled PCD that act in different cell types, tissues and organs within and across different monocot and dicot plant species.

Our morphological and histological analyses, as well as transcriptome profiling of the degenerating silk bases, points to PCD as a developmentally controlled process effectively bringing about silk base collapse. This tissue collapse has been described decades ago, and has been put forward as a possible mechanism to impede the pollen tube’s passage to the ovary in senescent silk strands (Bassetti and Westgate, 1993b) and to serve as a barrier for fungal infection (Snetselaar et al., 2001). In KIL1 loss-of-function plants, we observed a clear delay of the cell death and the subsequent tissue collapse at the silk base, coupled with an extension of the effective pollination period, whereas we noticed the opposite in KIL1 overexpression lines. While these results are in line with the idea that silk base degeneration terminates the effective pollination period in maize, it remains unclear how exactly PCD might be affecting pollen tube growth support and guidance. Target-oriented pollen tube growth was shown to depend on both long-range and short-range signals that are produced by sporophytic and female gametophytic tissues (Zhou and Dresselhaus, 2019). It is thus conceivable that viable TT cells in the silk strands are necessary to support pollen tube growth and provide guidance over the long distance between the receptive papilla silk hairs and the ovary.

Commercial maize hybrid seed production relies on successful pollination between a male pollen donor and a female plant. Synchronous pollen shed and rapid silk emergence typically results in an optimal seed set and efficient seed production. This requirement for synchronicity can limit production if a female inbred starts flowering too early with respect to male inbred pollen shedding. The typical pollen shedding window is 3–8 days, thus there is a practical limit to pollen availability (Fonseca et al., 2004; Liu et al., 2021). Female silk strands are optimally receptive to pollen shortly after silk emergence and their receptivity rapidly decreases in the days thereafter (Anderson et al., 2004). Male inbreds shedding after this interval are therefore often not considered for hybrid production. In addition, hybrids bred to be drought-tolerant often have an aggressive silking behavior under stress conditions that may be imparted by the female inbred parent (Edmeades et al., 2000; Liu et al., 2021). In optimal growth conditions, however, this type of female inbred often silks 4–5 days before pollen is available. In this specific case, increasing their silk receptivity could decisively improve hybrid seed production efficiency. Hence, increasing the silk receptivity window of this type of inbred would be advantageous to maximizing kernel set, improving cost of goods, and potentially opening new untapped genetic potential by allowing new genetic combinations with males with delayed shedding characteristics. However, penalties with regards to pathogen susceptibility might apply: Maize silks are susceptible for infection by the common smut pathogen Ustilago maydis for ∼8–14 days after emergence, and that susceptibility strongly decreases with progressing silk senescence (du Toit and Pataky, 1999). Likewise, the resistance of silks toward fungal ear rot caused by Fusarium graminearum increases after pollination-induced silk senescence, suggesting that silk senescence has a role in defense against pathogens (Valdivia et al., 2006). Therefore, future breeding efforts around the silk longevity trait will have to keep an eye on resistance toward the pathogens that infect maize silks.

Materials and methods

Pollination assays in the glasshouse and in the field

All plants in the glasshouse were grown from seeds in Jiffy-7 peat pellets (http://www.jiffypot.com). One week after planting, all seedlings were transferred to 7-l pots filled with peat-based soil with osmocote fertilizer (N.V. Van Israel, Belgium). Before silking, ears were covered with protective glassine bags to avoid self-pollination. Pollen collected from the pollen parents was manually applied to uncut silk strands at a single time point, which corresponded to the given DASE from the husk. To control for pollen viability, fresh pollen was collected in the morning from plants whose tassels had just started to shed pollen. Pollen from several plants was mixed, and 3 DASE and later time points were pollinated on the same days to control pollen viability via full kernel set in 3 DASE-pollinated ears. We only discarded entire pollination assays if the 3 DASE pollination showed poor kernel set, no other cobs were excluded.

Field trial setup at the Zwijnaarde field trial site

Kernels from the maize inbred line B104 were sown on a parcel located at the VIB Center of Plant Systems Biology trial site (51�00′35.1′N 3�42′57.0″E) with a sowing density of 133,333 plants per hectare, in the growing seasons of 2015 (May 12–June 10; three batches) and 2016 (May 4– May 28; two batches). In 2015, three sets were sown, without distributing plants into male (pollen) or female (seed) parents. In 2016, the first set included female (kernel) parents only (400 plants), while the second batch included male (pollen) parents (320 plants). Both the female and male parents were surrounded by protective “border plants.” Kernels were sown by hand in 10 adjacent rows of 5.6 m in length, 75 cm apart, and each containing 56 maize B104 plants. A total of 360 B104 plants were grown from all batches. Before silking, ears of female parents were covered with protective glassine bags to avoid self-pollination. Pollen collected from the male parents was manually applied to uncut silk strands at a single time point, which corresponded to given DASE.

Field trial setup at the Johnston field trial site

B73, B104, and PH1V69 inbred seeds were planted in a randomized complete block on Corteva Agriscience’s Johnston farm (Johnston, IA, USA) on May 13, 2017 and May 25, 2017. Four 4-row plot replicates for each inbred line were planted on the first planting date and a fifth replicate was planted with a 2-week delay to serve as pollen source for specific pollination dates. Pollination was conducted at 3, 7, 11, and 15 DASE by cutting silks ∼1 inch above the tip of the ear, and applying copious amounts of viable pollen directly afterward. The pollen was collected from either the same plant, delayed replicates, nursery rows on the Johnston farm, or greenhouse plants depending on timing of pollinations. Five to 10 ears per replicate were collected at the end of the season and both sides of each ear photographed. Kernel set in rings 6–10 were analyzed using an ImageJ (http://rsb.info.nih.gov/ij) macro for kernel detection with manual curation of each image to avoid counting the same kernels twice on images corresponding to the same ear.

Pollen tube germination density and efficiency of ovary fertilization

For pollen tube germination density analysis, exposed silk strands were pollinated at 3, 7, 11, and 15 DASE with pollen from Act1-Dpro:GUS reporter lines (Brettschneider et al., 1997). The most-apical 5 cm of the silks were collected and subjected to GUS staining. To detect GUS activity, silk strands were incubated in reaction buffer containing 0.5-M sodium phosphate buffer (pH 7.0), 0.5-mM ferricyanide, 0.5-mM ferrocyanide, 0.5% (v/v) Triton X-100, and 1-mM X-Gluc for 24 h in the dark at 37�C. Silk strands were mounted in 50% (v/v) glycerol and lengths were measured on scanned slides with ImageJ. Scoring of germinated pollen tubes was performed by using a DIC Olympus BX51 microscope. Pollen tube density was quantified as number of pollen tubes germinated on silk strands portion divided by the length of the corresponding silk region. One day (24 h) after pollination, ovaries from the mid-base section were fixed in 95% (v/v) acetone for 2 h. Afterward GUS staining was performed as described above; staining was monitored on ovary cross-sections under a Leica binocular microscope. To control for pollen viability, we performed crosses on control ears (3 DASE) and on senescent ears (7–15 DASE) in parallel.

Histological and microscopy analyses

CLSM of silk cross-sections was performed as follows: Silk strands harvested at 3, 7, 11, and 15 DASE were immersed in a fixative solution (4% [w/v] glutaraldehyde, 12.5-mM sodium cacodylate buffer pH 6.8), as previously described (Zhao et al., 2017). After 2 h, the fixative was removed and embedded in 50% (w/v) agar. Silk cross-sections (100 μm in thickness) were obtained with a vibratome (Leica VT1000S) and imaged with a ZEISS 710 CLSM using the natural autofluorescence of the organelle. Samples were imaged with an Olympus CLSM. To view autofluorescence, excitation was set to 488 nm and fluorescence was detected in the 500 nm to 600 nm range.

For Evans Blue staining, ∼1-cm long silk bases were incubated in 0.2% (w/v) Evans Blue solution at room temperature for 5 min and then washed extensively to remove excess dye and observed with a DIC Olympus BX51.

For TEM, maize silk strands from different time points (3, 7, 11, and 15 DASE) were cut into small pieces of 0.5 cm and immersed in a fixative solution (2.5% [w/v] glutaraldehyde, 4% [w/v] formaldehyde in 0.1-M Na-cacodylate buffer). Next, the cut pieces were placed in a vacuum oven for 30 min and then left rotating for 3 h at room temperature. The solution was later replaced with fresh fixative and samples were left rotating overnight at 4�C. After washing, samples were postfixed in 1% (w/v) OsO₄ with K₃Fe(CN)₆ in 0.1-M Na-cacodylate buffer, pH 7.2. Samples were dehydrated through a graded ethanol series, including a bulk staining with 2% (w/v) uranyl acetate at the 50% (v/v) ethanol step, followed by embedding in Spurr’s resin. To have a larger overview of the phenotype, semi-thin sections were first cut at a thickness of 0.5 �m and stained with toluidine blue. Ultrathin sections of a gold interference color were cut using an ultra-microtome (Leica EM UC6), followed by poststaining with uranyl acetate and lead citrate in a Leica EM AC20 and collected on formvar-coated copper slot grids. Ultrathin sections were imaged with a JEM 1400plus transmission electron microscope (JEOL) operating at 60 kV. The percentage of PCD nuclei was determined by counting over 100 nuclei per senescence stage. The nuclei were counted from the TTs of five to eight silks per plant; three to four plants were analyzed per senescence stage.

Silk transcriptome analysis by RNA-seq

For RNA-seq, the basal-most 2-cm sections of 25 silk strands from the mid-base region of four individual ears were pooled to obtain four biological replicates. Total RNA from silk material was extracted and purified with Spectrum Plant Total RNA Kit (Sigma-Aldrich, St Louis, MO, USA). RNA concentration and purity were determined on a Nanodrop spectrophotometer (Nanodrop Technologies Wilmington, DE, USA), and RNA integrity was confirmed on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Library preparation was conducted using the TruSeq RNA Sample Preparation Kit v2 (Illumina, San Diego, CA, USA). In brief, poly(A)-containing mRNA molecules were reverse transcribed, double-stranded cDNA was generated, and adapters were ligated. After quality control using the 2100 Bioanalyzer, clusters were generated through amplification using the TruSeq PE Cluster Kit v3-cBot-HS kit (Illumina) followed by sequencing on an Illumina NextSeq 500 instrument with the TruSeq SBS Kit v3-HS (Illumina) for the silk senescence time course, and an Illumina NextSeq 500 instrument for the KIL1-SRDX transcriptome analyses. Sequencing was performed in paired-end mode with a read length of 75 bp. Sequencing quality as well as read mapping and summarization were performed with a software pipeline on an in-house Galaxy server. Briefly, quality of raw data was verified with FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Next, quality filtering was performed using Trimmomatic as described (Bolger et al., 2014). Reads were subsequently mapped to version 3 of the maize Zea_mays_B73_v3_exome. Concordant paired reads that uniquely map to the genome were used for quantification on with Salmon (Patro et al., 2017). DEGs were identified with in Edge-R software package in R (Robinson et al., 2010) provided by the BioConductor database (http://bioconductor.org). Genes were considered as differentially expressed if their expression levels had an absolute Log₂(FC) > 1 and a FDR ≤ 0.05. Functional annotation of maize genes was derived from Ensembl Biomart (http://plants.ensembl.org/biomart/).

For MDS plots, the distance between each pair of samples was calculated based on the 500 genes with the largest FCs. GO enrichment analyses were carried out with the PANTHER software on the GO resource website (http://geneontology.org). The Hierarchical Clustering Analysis was performed with the Orange Canvas software (http://orange.biolab.si/) using Euclidean Distance as the distance measure and the average linkage clustering option. The Euclidean Distances were calculated between all 28 datasets for 39,479 genes and then normalized. The resulting matrix was used to build both the dendrograms and heatmap. The heatmap in Supplemental Figure S4C representing the expression value in the different RNA-seq samples of the genes with a FDR < 0.05 and a Log₂(FC) > 2 or less than −2 in at least one of the comparison over time was drawn with the heatmapper online software (http://www.heatmapper.ca/) with the expression value normalized by row.

K-means clustering in Supplemental Figure S4E was performed on the Log₂ values of the expression of the DEGs over the silk senescence time course with the WebMeV software https://mev.tm4.org). The parameters used were 12 clusters as output, Pearson’s coefficient of correlation as a distance metric, and 50 as the maximum number of iterations.

To analyze SILK1 expression in different maize tissues, average expression values were generated from a proprietary database based on multiple unrelated RNA-seq datasets created by Corteva Agriscience. Up to 468 samples, nearly all from B73, were collected from different tissues, at different developmental stages, under different growing conditions and treatments, and profiled using Illumina HiSeq RNA profiling technology over multiple years, creating a proprietary expression atlas. Sequences were mapped to a proprietary B73 genome draft (comparable to B73 RefGen version 2.0 in quality), and expression levels were analyzed and normalized by the RSEM method (Li and Dewey, 2011), and expressed as “Reads Per Kilobase of transcript per ten million Mapped reads.”

Nicotiana benthamiana infiltration

The coding sequences of KIR1, KIL1, and the control TF gene TMO5 were cloned under the control of the CaMV 35S promoter in the bicistronic GFP-positive destination vector pB7WG2D,1. Nicotiana benthamiana plants were grown on soil (Saniflor V4398609) for 4–6 weeks before infiltration with a suspension of transformed Agrobacterium (Agrobacterium tumefaciens) strain LBA4404 cells resuspended at a cell density (as determined by OD₆₀₀) of 1.5 in infiltration medium (10-mM MgSO₄, 10-mM MES pH 5.8, 100-μM acetosyringone).

Protoplast preparation, transfection, transient expression assays

Protoplasts were isolated and transient expression assays were carried out as described previously (Simmons et al., 2020), with some modifications as described below. B73 plants were grown for 4 days under controlled environmental conditions under a 16-h light/8-h dark photoperiod or under continuous light at 24�C, then moved to the dark chamber for 7 days. Mesophyll protoplasts were isolated from the middle part of second leaves. Protoplasts were co-transfected with 10 mg of a reporter plasmid that contains the firefly luciferase (fLUC) reporter gene driven by the corresponding promoter, 2 mg of normalization plasmid expressing Renilla LUC (rLUC) under the control of the 35S promoter and 10 mg of the effector construct. The total amount of DNA was equalized in each experiment with the p2GW7-GUS mock effector plasmid. After transfection with polyethylene glycol for 5 min, protoplasts were incubated overnight in the dark with gentle agitation at room temperature and then lysed; fLUC and rLUC activities were determined using the Dual-Luciferase reporter assay system (Promega, Madison, WI, USA). Variations in transfection efficiency and technical errors were corrected by normalization of fLUC activity by that for rLUC. The mean values were calculated from five measurements and each experiment was repeated at least 3 times.

Flow cytometry screening assay

Maize mesophyll protoplast suspensions were co-transfected with: (1) a fluorescent reporter, and a control plasmid (35Spro:nlsGFP + 35Spro:GUS) and (2) a fluorescent reporter, and either KIR1 or a death-inducing TF gene (35Spro:nlsGFP + 35Spro:KIR1 or 35Spro:candidateTF). To enhance throughput and reproducibility, transfections were performed as automated assays in 48-well plates as described (Vanden Bossche et al., 2013). Sixteen hours after transfection, protoplast suspensions of 30 wells representing one sample were pooled together, centrifuged at 700 g at room temperature for 3 min, and supernatant was carefully removed. Sedimented protoplasts were dissolved in 600 μL of MMg buffer (Simmons et al., 2020) and divided into four replicates (150 μL each). Two replicates were evaluated by means of flow cytometry on Day 1. The other two replicates were kept in the dark at 16�C and subjected to flow cytometry on Day 4, both using a FACSAria (BD Biosciences, Franklin Lakes, NJ, USA) instrument using 488-nm excitation and measuring emission at 530/30 nm for green fluorescence. Data were analyzed using FlowJo software (http://www.flowjo.com/).

Gene expression induction experiments in A. thaliana

Inducible gene expression was carried out using an estradiol-based inducible gene expression system (Siligato et al., 2016). Five-day-old seedlings were transferred to plates containing 10-μM estradiol as previously described (Gao et al., 2018).

RT-qPCR

An amount of 1-μg purified total RNA was subjected to first-strand cDNA synthesis with the iScript cDNA synthesis kit (BioRad, Hercules, CA, USA). The resulting cDNA was dissolved in ultra-pure water and mixed with the LightCycler 480 SYBR Green I Master (Roche, Basel, Switzerland) and 0.5-μM gene-specific primers (Supplemental Table S7; Gao et al., 2018) using JANUS automated pipetting station (Perkin Elmer, Waltham, MA, USA). qPCR was performed with a LightCycler system (Roche) in a program of 45 cycles and a primer annealing temperature of 60�C. All PCRs were performed in triplicate. The 2(Delta Delta C(T)) method was used to analyze the relative changes in gene expression. Relative expression levels were first normalized to a reference gene (ACTIN for silk samples; PEX4 for seedlings) and then to the respective expression levels in the WT.

Cloning and preparation of transgenic lines

All constructs (TF candidates and promoters) used in this study were generated by recombining synthetic DNA fragments synthetized by GeneScript (https://www.genscript.com; either attL4_att1R or attL1_attL2-flanked) into the destination vector pBbm42GW7 (gatewayvectors.vib.be).

KIL1-OE and KIL1-SRDX constructs were obtained by GreenGate cloning (Lampropoulos et al., 2013). A putative promoter region of 1,491-bp upstream of the start codon of GRMZM2G145461 was synthetized by GeneScript (https://www.genscript.com) and cloned into the pGGA000 entry vector. The coding sequences of KIL1 and KIL1-SRDX (SRDX = 5′-CTCGATCTGGATCTAGAACTCCGTTTGGGTTTCGCTTGA-3′) were synthetized by GeneScript and cloned into pGGC000. SILK1pro:KIL1 (KIL1-OE) and SILK1pro:KIL1-SRDX (KIL1-SRDX) constructs were generated by GreenGate cloning with the entry plasmids described and the pGGBb-AG destination vector (gatewayvectors.vib.be). Destination vectors carrying constructs of interest were transformed into heat shock-competent Agrobacterium strain EHA101, which was then used for B104 embryo transformation, as published (Coussens et al., 2012).

For the reporter constructs used in the transient expression assays, 1.5-kb promoter fragments upstream of the start codon of PCD-associated genes were cloned into pDONR-P4-P1R to generate pEN-L4-Pro-R1 vectors. Promoters of ZmRNS3a, ZmRNS3b, and ZmMC9 with attB4 and attBR1 borders were synthesized by GenScript, and promoters of ZmDMP4a, ZmDMP4b, ZmEXI1, and ZmBFN1 were PCR amplified from genomic DNA with primers listed in Supplemental Table S7. pEN-L4-Pro-R1 vectors were recombined with pEN-L1-fLUC-L2 by Multisite Gateway LR cloning with pm42GW736. For the generation of effector constructs (that were also used in the flow cytometry assay), the pEN-L1-ORF-R2 plasmids (with the coding sequences of KIR1, KIL1-5, and other candidate TF genes) were used to introduce the coding sequences by Gateway LR cloning into p2GW7 for overexpression.

All transgenic lines were backcrossed once. Five independent KIL-OE lines and 11 KIL1-SRDX lines were generated, and the T₀ generation was prescreened for silk phenotypes. Pollination assays were performed on the T₂ and T₃ generations (still segregating) with two KIL-OE and three KIL-SRDX lines.

CRISPR-based targeted mutation of KIL1

Two single-guide RNA (sgRNA) sequences 5′-ATCGAGTCCACATTGCCGCC-3′ (KIL1-qRNA1) and 5′-CCCATGGCTCTCGCGCGTGC-3′ (KIL1-qRNA2), were selected to target the region of KIL1 encoding the N-terminal NAC domain. Predicted efficiencies, specificities and off-targets of sgRNAs were evaluated by CRISPR-OR analysis (http://crispor.tefor.net) using the B73 v4 maize reference genome. pCBC-MT1T2 (Xing et al., 2014) was used as a template for cloning two sgRNAs and cloned in CRISPR vector pBUN411-Sp (Pedroza-Garcia et al., 2021). The rice U3 promoter was used to drive sgRNA expression. DNA sequencing of the sgRNA target sites in exon 1 identified one T₀ plant with a single-nucleotide insertion of one adenine. One TIDE primer pair that flanked both target sites was used to detect the mutation. Primer sequences can be found in Supplemental Table S7. This T₀ transformant was already homozygous for the kil1 mutation, and was thus not backcrossed but self-pollinated. Homozygous, Cas9-free T₂ plants were used for the first pollination assay; the second pollination assay was performed on the T₃ generation.

Data availability statement

The RNA-seq data sets in this study are available in the ArrayExpress (https://www.ebi.ac.uk/arrayexpress; accession number E-MTAB-10850). All other data generated or analyzed during this study are included in this published article and its Supplemental Information files. Sequence data from this article can be found in Tables�1 and 2.

Supplemental data

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

Supplemental File S1. Multiple protein alignment used for the phylogenetic tree shown in Figure�3C.

Supplemental Files S2. Newick format for the phylogenetic tree shown in Figure�3C.

Supplemental Figure S1. Kernel set at rings 6–10 after manual synchronous pollination in field conditions.

Supplemental Figure S2. The base of unpollinated senescent silk shows subcellular hallmarks of PCD.

Supplemental Figure S3. Pollen tube density gradually decreases during silk senescence.

Supplemental Figure S4. Transcriptome analysis reveals differentially regulated genes during silk senescence.

Supplemental Figure S5. KIL1 induces cell death in Arabidopsis.

Supplemental Figure S6. Expression of SILK.

Supplemental Figure S7. KIL1 depletion delays silk senescence.

Supplemental Figure S8. KIL1 overexpression reduces cob size and activates expression of cell death-associated genes.

Supplemental Figure S9. Overexpression of KIL1 and loss of KIL1 function reduces or increases kernel set after late pollination of independent KIL transgenic lines.

Supplemental Table S1. Number of reads and percentage mapped on the V3 version of the maize genome for each RNA-seq sample.

Supplemental Table S2 . Top 10 most enriched GO terms for the DEGs at 15 DASE compared to 3 DASE.

Supplemental Table S3. Expression value and expression comparison for the different stages of silk senescence for a list of the closest orthologs of Arabidopsis dPCD genes that show an increase in expression during silk senescence.

Supplemental Table S4. Number of reads and percentage mapped on the V3 version of the maize genome for each RNA-seq sample.

Supplemental Table S5. GO enrichment analysis.

Supplemental Table S6 . Expression value and expression comparison between the WT and transgenic lines for the closest orthologs of Arabidopsis dPCD genes that show a significant increase in expression during silk senescence.

Supplemental Data Set 1. List of DEGs after silk emergence.

Supplemental Data Set 2. TF-encoding genes with increasing expression levels throughout senescence and more than 100 read counts per million at 15 DASE.

Supplemental Data Set 3. Expression level of SILK1 and KIL1-6 genes in Corteva Agriscience’s proprietary B73-specific gene expression database.

Supplemental Data Set 4. List of DEGs in WT, KIL1-OE, and KIL1-SRDX after silk emergence.

Supplemental Data Set 5. Summary of statistical analyses.

 

M.S., A.D., N.B., J.E.H., and M.K.N. designed the study. M.S. and A.D. performed most of the research and analyzed data. C.R.S. analyzed bioinformatic data. F.D.W. performed and analyzed dual luciferase assays. N.D. performed and analyzed tobacco infiltrations and analyzed bioinformatic data. N.S. analyzed the RNA sequencing data sets. M.C.R. contributed to pollination assays, RNA sequencing and analysis. L.Z. and T.D. assisted with the pollen tube growth assays. R.D.R. performed and analyzed transmission electron microcopy. S.A. and L.P. generated transgenic maize lines. M.K.N., M.S., and A.D. wrote the manuscript with input from all the other authors.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plcell) is: Moritz K. Nowack (moritz.nowack@vib.be).