Maize YABBY genes drooping leaf1 and drooping leaf2 regulate floret development and floral meristem determinacy

A major goal in plant biology is to understand the factors that regulate meristem activity. Meristems, which are active, pluripotent stem cell tissues, produce all postembryonic organs of flowering plants (Greb and Lohmann, 2016). Meristem determinacy (degree of meristem activity) is a crucial factor that shapes vegetative, inflorescence and floral architectures. Vegetative and inflorescence meristems are indeterminate, producing an unspecified number of lateral primordia. Floral meristems (FMs) are generally determinate, initiating a set number of floral whorls and organs before undergoing terminal differentiation. Commonly, eudicot flowers are composed of four whorls of floral organs (outermost to innermost: sepal, petal, stamen and carpel). Similarly, in the monocots grass florets (flowers), including those in maize (Zea mays), are arranged in whorls of floral organs, some of which have grass-specific names (outermost to innermost: lemma, palea, stamen and carpel). As each grain is the product of one floret, regulation of FM activity is a key agronomic trait.

FMs pattern flowers through the combinatorial activity of three classes of gene functions that dictate organ identity and FM determinacy (Coen and Meyerowitz, 1991; Pelaz et al., 2000). In Arabidopsis thaliana, carpels are specified by the MADS-box transcription factor AGAMOUS (AG) (Yanofsky et al., 1990). AG is expressed in the FM and controls FM determinacy by repressing expression of the stem cell regulator WUSCHEL (Lenhard et al., 2001; Lohmann et al., 2001). In the cereals, AG orthologs have expanded and undergone subfunctionalization during grass evolution, leading to redundancy in the regulation of FM determinacy (Mena et al., 1996; Yamaguchi et al., 2006; Dreni et al., 2011). The maize AG ortholog zea agamous1 (zag1) imposes FM determinacy with less obvious roles in regulating floret organ identity as supernumerary carpels develop in pistillate florets of zag1 mutants (Mena et al., 1996). ZAG1 interacts physically with the AG-LIKE6 (AGL6) subfamily member bearded ear (BDE; also known as ZAG3), and zag1; bde double mutants reveal a synergistic interaction in regulating FM determinacy (Thompson et al., 2009). Pistillate and staminate FMs are more indeterminant in the maize indeterminate floral apex1 (ifa1) mutant, and ifa1 interacts synergistically with zag1 to regulate FM determinacy (Laudencia-Chingcuanco and Hake, 2002). In the bisexual florets of rice (Oryza sativa), AG orthologs OsMADS3 and OsMADS58 regulate floret organ identity and FM determinacy, respectively (Yamaguchi et al., 2006; Dreni et al., 2011).

The Arabidopsis YABBY family member CRABS CLAW (CRC) is required for proper growth of the gynoecium (Alvarez and Smyth, 1999). Loss-of-function mutations in CRC consistently reduce stylar growth and result in incomplete medial fusion of carpels. crc mutants occasionally produce three carpels compared with two in wild type, suggesting that CRC is necessary to promote FM determinacy (Alvarez and Smyth, 1999; Bowman and Smyth, 1999). Expression of CRC is restricted laterally to developing carpels and nectaries (Bowman and Smyth, 1999). The rice CRC ortholog DROOPING LEAF (DL) is required for carpel identity, as carpels undergo homeotic transformation to stamens in strong loss-of-function alleles of dl mutants (Yamaguchi et al., 2004). Transformed stamens are variable in number, indicating that DL also regulates FM determinacy. DL is expressed in carpel primordia of rice florets (Yamaguchi et al., 2004). Genetic analysis indicates that DL and the rice AGL6 subfamily member MOSAIC FLORAL ORGAN 1/OsMADS6 redundantly regulate FM determinacy (Li et al., 2011).

Here, we report that the maize CRC co-orthologs drooping leaf1 (drl1) and drl2 are required for the development of dimorphic, unisexual ear and tassel florets. drl1 floret phenotypes and FM indeterminacy are enhanced by natural variant and transposon alleles of drl2. The drl paralogs are co-expressed in lateral organ primordia initiated by the FM, but not within the FM. Gene regulatory network (GRN) inference, genetic interaction and expression analyses suggest that DRL1 and ZAG1 target each other and a common set of downstream genes that function during floret development. Our results demonstrate that the drl genes are required for floret patterning, to impose meristem determinacy via some non-cell-autonomous mechanism, and, together with zag1, define a regulatory module that provides crucial control of floret patterning and FM determinacy in the development of grain-producing structures.

Maize staminate and pistillate florets are produced on the tassel and ear, respectively (Kiesselbach, 1949). During tassel and ear development, branching events from multiple meristem types (Irish, 1997) ultimately give rise to floret whorls housed in grass-specific spikelets (Clifford, 1987). The indeterminate inflorescence meristem (IM) of the tassel and ear initiates determinate spikelet pair meristems (SPMs); additionally, in the tassel the IM initiates indeterminate branch meristems (BMs). Each SPM produces a pair of determinate spikelet meristems (SMs), each of which gives rise to two glumes. Afterwards, each SM initiates a lower floral meristem (LFM) and then converts identity to an upper floral meristem (UFM). Each determinate LFM and UFM gives rise to a lemma, a palea, two lateral-abaxial lodicules, three stamens and three carpels (Cheng et al., 1983). Two lateral-adaxial stamens are spaced widely relative to the medial-abaxial stamen (Irish et al., 2003). Three connately fused carpels form the single pistil; however, only the two lateral-abaxial carpels (indeterminate carpels, C_i) form an elongated silk, whereas growth of the medial-adaxial carpel (determinate carpel, C_d) is limited in growth to envelop the single ovule (Randolph, 1926; Bonnett, 1953). The ovule consists of a mostly enclosing inner and a partially enclosing outer integument plus nucellar tissue that contains the embryo sac, and fills the locule formed by the three fused carpels (Kiesselbach, 1949). After organ initiation, sex determination in the tassel and ear culminates in abortion of the carpel whorl in staminate florets and arrest of stamen primordia in pistillate florets, respectively (Cheng et al., 1983) (Fig. 1A,B).

Likely null mutations of drl1 displayed aberrant pistillate and staminate floret morphologies. Macroscopically, drl1 mutant ears were sterile, with underdeveloped silks consisting of reduced, unfused carpel walls that failed to enclose an expanded nucellus-like structure (Fig. 1C,D). drl1 pistillate phenotypes were reminiscent of the floret phenotypes described for the ifa1 mutant (Laudencia-Chingcuanco and Hake, 2002). We found drl1 and ifa1 to be allelic through genetic noncomplementation of mutant alleles (Fig. S1) and by sequencing the drl1 locus in ifa1 mutant plants (Strable et al., 2017). drl1 and its paralogous genetic enhancer locus, drl2, encode CRC co-orthologs in the YABBY family of transcriptional regulators (Strable et al., 2017). Genetic combinations between drl1 alleles and the loss of- or low-function drl2-Mo17 natural variant allele (hereafter referred to as drl2-M) or the strong drl2-DsD08 transposon allele (Strable et al., 2017) enhanced all aspects of the drl1 floret phenotype, such that florets from double mutants displayed multiple, expanded nucellus-like structures that appeared to originate from sustained FM activity in the upper floret (UF) (Fig. 1C,D; Fig. S2). The synergistic genetic interactions between drl1 and drl2 mutant and natural variant alleles in florets were dose sensitive, consistent with dosage effects observed for vegetative traits (Strable et al., 2017). In an F₂ population with varied dosage of drl2-M, florets of drl1-R; drl2-M/+ plants were intermediate in severity between drl1-R homozygotes and drl1-R; drl2-M double homozygotes (Fig. 1C). Collectively, these observations suggest that the drl loci regulate pistillate floret development in a dose-dependent manner.

In the drl1 mutant tassel, we observed an ectopic stamen periodically in the UF of sessile [3.14±0.06 (mean±s.e.m.)] and pedicellate (3.05±0.04) spikelets, compared with three stamens in normal spikelets (Fig. 1E-G). Histological examination of mature drl1 mutant spikelets revealed that the infrequent extra stamen originated internal to the normally placed and numbered lemma and palea in the outer whorl (Fig. 1F). FM indeterminacy was enhanced in drl1; drl2 double mutants, in which stamen number was increased in both the UF and lower floret (LF) of sessile [4.49±0.09 (P=1.0×10⁻¹⁹) and 3.41±0.08 (P=4.6×10⁻⁶), respectively] and pedicellate [3.92±0.11 (P=3.0×10⁻¹⁰) and 3.19±0.08 (P=0.015), respectively] spikelets (Fig. 1E-G). Such differences were significant between stamen number in the UF and LF within sessile and within pedicellate spikelets (P<10⁻⁶), and for UFs, between sessile and pedicellate spikelets (P<10⁻³) (Fig. 1G). These data suggest that the drl genes participate differentially in determinacy pathways of upper and lower staminate FMs. An alternative explanation is that the drl genes function equally in UFM and LFM determinacy pathways, but that LFM determinacy is more hardwired relative to the UFM, which is therefore perhaps more sensitized to loss of drl gene function. Finally, the observations also suggest that ectopic stamens originate from sustained activity of the mutant FM.

Some floret phenotypes were specific to drl1; drl2 double mutants. We observed an ectopic primordium with lodicule-like cellular morphology and vascularization occasionally in the position of a presumptive, suppressed adaxial-medial lodicule in the UF (Fig. 1F, right panel, arrowhead), indicating a possible role for drl gene products in imposing zygomorphy (Irish et al., 2003; Bartlett et al., 2015). We also observed macrohair-like structures along the apical ridge of drl1-R; drl2-M supernumerary anthers (Fig. S3). Macrohair production is generally limited to the adaxial epidermis of the adult leaf blade and is frequently used as a morphological marker for leaf polarity (Juarez et al., 2004). Though these ectopic structures were infrequent, they lacked the multicellular bases of leaf blade macrohairs (Becraft and Freeling, 1994) and were consistently associated with supernumerary anthers with altered morphology. Such amorphic anthers had aberrant theca that lacked pollen sacs and were often fused to morphologically normal anthers. These data suggest that the drl genes are necessary for complete suppression of macrohair formation on reproductive floral organs, similar to AG and SHATTERPROOF1/2 in suppressing trichome initiation on floral organs in Arabidopsis (Ó'Maoiléidigh et al., 2018). Alternatively, the ectopic structures may be a mosaic of anther and leaf identities due to a partial loss of stamen identity in drl1; drl2 double mutants.

We tracked the developmental basis of drl1 and drl1; drl2 mutant phenotypes in mid- and later-staged pistillate florets with scanning electron microscopy (SEM). Prior to sex determination, the inner whorl of normal UFs consisted of a medial-adaxial C_d primordium (determinate carpel) and two lateral-adaxial C_i primordia (indeterminate carpels), all of which were connately fused (Fig. 2A). This gynoecial whorl was flanked by a whorl of three pre-degenerate stamen primordia. The LF lagged in development, with an FM and recently initiated stamen primordia (Fig. 2A). In mid-staged UFs of drl1 mutants, the medial-adaxial C_d initiated with extreme delay or, often, not at all (Fig. 2B,C). Carpel walls did not fuse entirely in drl1 mutants, and development of integument tissues appeared to be compromised (Fig. 2B,C), resulting in the eventual single protruding nucellus-like structure observed in mature drl1 mutant florets (Fig. 1D). drl1 mutant UFs had extra whorls of lateral C_i (Fig. 2C); however, shifts in phyllotaxis between each extra whorl complicated assigning ab- or adaxial orientation relative to the palea axil. The medial-adaxial C_d was similarly greatly reduced or suppressed in mid-staged UFs of drl1; drl2 double mutants, yet the double mutants displayed multiple whorls of lateral C_i indicating prolonged FM activity (Fig. 2D-F). Additionally in the axil of each C_i whorl of drl1; drl2 double mutants, we frequently observed an ectopic structure that we interpreted to be a lodicule-like and/or anther-like primordium based on position and morphology (Fig. 2D-F, asterisks).

In normal later-staged UFs, lateral-adaxial C_i primordia appeared paired and elongate, whereas the reduced medial-adaxial C_d was a ridge of cells just prior to enveloping of the ovule (Fig. 2G). Multiple whorls of paired lateral-adaxial C_i primordia were obvious in similarly staged drl1 mutant UFs (Fig. 2H), whereas lateral-adaxial C_i primordia observed in later-staged UFs of drl1; drl2 double mutants indicated the presence of an extra, intra-whorl fused or partially fused C_i primordium (Fig. 2I). Additionally, we often detected involution of the palea, or, alternatively, partial fusion of paleas, along the medial axis in drl1; drl2 double mutant florets (Fig. 2E,I, arrowheads), which may indicate crowding within the inner whorl of the floret or ectopic palea that initiate within their normal whorl. Taken together, these observations indicate that the drl genes are required for proper patterning of pistillate florets, including C_d elaboration or initiation, and to impose FM determinacy.

To examine the temporal and spatial patterns of drl transcript accumulation during inflorescence and floret development, we performed RNA in situ hybridization. In median longitudinal sections of the developing ear, drl1 transcripts were detected in the IM periphery, which spatially corresponds to cryptic bract anlagen (Whipple et al., 2010) (Fig. S4A). drl1 transcripts continued to accumulate in outer glume primordia (Fig. 3A), but not in the SM, as marked by accumulation of knotted1 (kn1) transcripts (Jackson et al., 1994) (Fig. 3B). The accumulation pattern of drl1 transcripts persisted in lateral organs of later-staged SMs where they were detected in lemma and palea primordia (Fig. 3C,D, Fig. S4B), expression patterns that were also observed for drl2 (Fig. S4C). In more advanced pistillate florets, drl1 transcripts accumulated in carpel primordia that had initiated in the UF and LF, but not in the central presumptive ovule primordium of either floret (Fig. 3E,F). In developing staminate florets, drl1 transcripts accumulated similarly in lateral primordia that were initiated by the FM, but not within the FM (Fig. 3G,H). drl expression dynamics across developing inflorescences were supported using publicly available transcriptomic data (Fig. S4E; www.maizeinflorescence.org). To summarize, the drl genes were expressed in cryptic bracts, in lateral organ primordia initiated by the SM (glumes), and in primordia of outer (lemma and palea) and inner (carpels) whorl organs initiated by the FM. drl expression in carpel primordia correlated with the organs for which development was altered in drl1 and drl1; drl2 mutant florets. However, the indeterminate FMs observed in drl mutant florets are best explained by mis-regulation of FM activity, yet drl expression was limited to organs derived from the meristems and was excluded from the meristem. These points strongly suggest that drl regulates meristem activity via a non-cell-autonomous mechanism. Consistent with this hypothesis, drl1 and drl1; drl2 mutants also display a dose-dependent reduction in vegetative shoot apical meristem (SAM) size even though the drl genes are expressed in leaf primordia and not in the SAM proper (Strable et al., 2017).

To interpret the indeterminant drl mutant pistillate florets further, we examined the expression patterns of the FILAMENTOUS FLOWER homolog zea zyb15 (also known as yabby15 or yabby8) (Strable et al., 2017) and kn1 (Jackson et al., 1994) in the drl1-R; drl2-M/+ background. In the inflorescence, zyb15 is expressed in cryptic bract (Whipple et al., 2010) and outer whorl primordia of florets (Gallavotti et al., 2011). We observed zyb15 transcript accumulation in glume, lemma, palea and carpel primordia, but not in the FM or in stamen primordia, for both normal and drl1-R; drl2-M/+ developing pistillate UF and LF (Fig. 3I-N). Interestingly, zyb15 transcript accumulation persisted longer in glume primordia compared with drl1 accumulation (Fig. 3M, compare with 3E). The kn1 gene is expressed in meristematic cells and is downregulated in cells recruited to form a lateral domain on the flank of the meristem and in lateral organ primordia (Jackson et al., 1994). We observed that kn1 transcripts were absent from normal later-staged pistillate UFs that had undergone terminal differentiation to an ovule primodium, whereas in similarly staged drl1 mutant UFs, kn1 transcript accumulation persisted throughout the gynoecial axis, demonstrating that drl1; drl2 FMs are more indeterminate (Fig. 3O).

In maize, FM determinacy is regulated redundantly by zag1 and bde genes, both of which are expressed dynamically throughout floret development, including in the FM, and whose encoded proteins physically interact (Schmidt et al., 1993; Mena et al., 1996; Thompson et al., 2009). Additionally, the ifa1 allele imposes FM determinacy redundantly with zag1 (Laudencia-Chingcuanco and Hake, 2002). Similarly, we observed extreme loss of determinacy in pistillate florets of zag1-mum1; drl1-R; drl2-M triple mutants, in which floret axes displayed iterative secondary and tertiary branch-like lateral growth from the axils of ectopic palea or bracts (Fig. 4A-E).

To gain insight into a potential regulatory module and shared targets for DRL1 and ZAG1, we mined an integrated atlas of gene expression, protein abundance, and regulatory networks generated from multiple tissues, including ear florets, across maize development (Walley et al., 2016). From this atlas, Walley and co-workers demonstrated that integrating transcriptome, proteome and phosphoproteome datasets into unified GRNs significantly improved the predicative power of the GRN. drl1, drl2 and zag1 mRNAs and their encoded non-modified proteins and phosphoproteins accumulated differentially throughout pistillate floret development (Fig. 4G). When classified as regulators in the integrative transcriptome, proteome and phosphoproteome GRN, DRL1 and ZAG1 transcription factors shared 51.4% of their target genes (Fig. 4H), implying that DRL1 and ZAG1 co-regulate many genes to control floret development. Among the high-confidence edge scores, DRL1 is predicted to target seven of the 13 YABBY family members, including itself and drl2 (Fig. 4I, column 7), indicating potential auto- and cross-regulation activities at the drl loci, a hypothesis that is supported genetically by dosage-related phenotypes in drl1 and drl2 mutant ears (Fig. 1C). The seven YABBY target genes are expressed throughout inflorescence development, and a majority of them show peak expression levels in late-staged ears and tassels, which parallels floret development (Fig. 4I, ear and tassel columns). Furthermore, drl1, drl2 and yab5 are shared predicted high-confidence target genes with ZAG1 (Fig. 4I). We next looked at high-confidence MADS box target genes predicted to be shared between DRL1 and ZAG1 regulators and found 26 genes that are potentially co-regulated and are co-expressed during inflorescence development (Fig. 4I). Among the co-regulated target genes are zag1 and bde, which have been shown previously to have roles in floral development and meristem determinacy (Mena et al., 1996; Thompson et al., 2009). The extreme loss of determinacy in zag1-mum1; drl1-R; drl2-M triple mutant pistillate florets (Fig. 4A-E) and ectopic zag1 transcript accumulation in late-staged drl1-R; drl2-M double mutant ears (Fig. 4J) supports a putative complex relationship between DRL1-ZAG1 regulators and their drl1, drl2 and zag1 targets. DRL1 and ZAG1 are predicted to target silky1 (si1) (Fig. 4I), the maize APETALA3/DEFICIENS ortholog, which is required for lodicule and stamen identity, and is expressed in developing lodicule and stamen primordia of staminate and pistillate florets (Ambrose et al., 2000; Chuck et al., 2008; Bartlett et al., 2015). We found si1 to be mis-expressed at earlier developmental stages in drl1-R; drl2-M double mutant ears compared with its late-stage expression in ears from normal siblings (Fig. 4J). We explored this result by RNA in situ hybridization in drl1-R; drl2-M/+ ears and found si1 transcript accumulation marked ectopic primordia in the UF and was strongly expressed throughout the LF (Fig. S5). By comparison, in normal pistillate UFs and in the LFs si1 expression was restricted to degenerating stamen primordia and lodicules. The mis-expression of si1, together with the appearance of ectopic lodicule-like and/or anther-like structures in axil of each Ci whorl of drl1; drl2 double mutants (Fig. 2), suggests that the drl genes are necessary to promote and/or maintain boundary identity between their expressed whorl 4 and adjacent floret whorls by suppressing si1 expression.

We uncovered floral-expressed genes that had been characterized previously but not described as putative targets of DRL1 and ZAG1. These candidate genes include the squamosa promoter-binding transcription factor-encoding teosinte glume architecture1 (Wang et al., 2005) and the ASYMMETRIC LEAVES 2 (AS2) homologs indeterminate gametophyte1 (ig1) and ig1-as2 like1 (Evans, 2007). Other candidate DRL1 and ZAG1 target genes are required for the development of dimorphic, unisexual ear and tassel florets, such as anther ear1, which encodes a gibberellin biosynthetic enzyme necessary to promote stamen abortion in pistillate florets (Bensen et al., 1995), grassy tillers1, which encodes a homeodomain leucine zipper transcription factor required to repress carpel growth in staminate florets (Whipple et al., 2011), and nana plant1 (na1), which encodes a brassinosteriod biosynthetic enzyme that represses carpel growth in the center whorl of staminate florets (Hartwig et al., 2011). We tested the relationship between na1 and drl genes through genetic interaction analysis. Pistillate florets of na1-R; drl1-R double mutant ears ranged from a single large nucellus subtended by ectopic pistils to reduced florets with many ectopic pistils (Fig. 4F). We observed normal, unisexual staminate florets in na1-R tassels for both greenhouse and field conditions (Fig. S6), suggesting that the tasselseed phenotype is specific to allele, genetic background and/or environmental condition.

Genetic, expression and evolutionary analyses indicate that CRC and orthologous genes are key regulators of floral development across diverse angiosperms (Alvarez and Smyth, 1999; Bowman and Smyth, 1999; Eshed et al., 1999; Yamaguchi et al., 2004; Fourquin et al., 2005, 2014; Orashakova et al., 2009; Nakayama et al., 2010; Yamada et al., 2011; Pfannebecker et al., 2017). Our results demonstrate crucial roles for the maize drl genes in regulating stem cell homeostasis and patterning of inner-whorl organs in dimorphic, unisexual florets. Unexpected findings of this study were the presence of ectopic primordia with lodicule-like cellular morphology, vascularization and medial placement in staminate florets (Fig. 1F), and, in the axils of carpel primordia in pistillate florets, the development of ectopic structures that we interpreted to be lodicule-like or anther-like primordia based on morphology and position (Fig. 2). These observations, together with the mis-expression of si1 in drl1; drl2 double mutant ears (Fig. 4J, Fig. S5), strongly hints at a possible antagonistic relationship between drl genes and the B-class MADS box gene si1 that specify lodicule (whorl 3) and stamen (whorl 2) identities (Ambrose et al., 2000; Bartlett et al., 2015). Understanding the genetic relationship between the drl genes and B-class genes si1 (Ambrose et al., 2000) and sterile tassel silky ear1 (Bartlett et al., 2015) may help clarify mechanisms that underlie their function in floret patterning and potentially in regulating developmental programs that control zygomorphy.

We hypothesize that the drl gene products function non-cell-autonomously in or through pathways that signal from lateral primordia, through boundary domains, to regulate developmental programs that impose FM determinacy. Typically, FMs terminate upon correct spatiotemporal initiation of all floral organ primordia. In Arabidopsis, cessation of FM activity is concurrent with the expression of AG, which integrates stem cell homeostasis with floral patterning pathways; AG directly represses WUS expression and subsequently FM activity is lost (Lenhard et al., 2001; Lohmann et al., 2001). AG indirectly represses WUS expression by directly activating the expression of KNUCKLES (KNU), which encodes a C2H2 zinc-finger transcription factor that directly represses WUS (Payne et al., 2004; Sun et al., 2009). KNU (Sun et al., 2009, 2014) and CRC (Gómez-Mena et al., 2005; Ó’Maoiléidigh et al., 2013) are direct targets of AG and function in parallel to regulate FM determinacy (Yamaguchi et al., 2017). crc; knu double mutants display a synergistic interaction with a highly indeterminate floral axis (Yamaguchi et al., 2017). Recently, CRC and CRC-AG were shown to impose FM determinacy by controlling auxin homeostatic (Yamaguchi et al., 2017) and biosynthetic (Yamaguchi et al., 2018) pathways, respectively. However, whereas WUS, AG and KNU expression domains overlap spatially and temporally in the FM during floral development (Lenhard et al., 2001; Lohmann et al., 2001; Sun et al., 2014), CRC is expressed in adjacent lateral carpel primordia at slightly later developmental stages (Bowman and Smyth, 1999), suggesting additional factors may provide requisite spatiotemporal inputs (Goldshmidt et al., 2008). Unlike in Arabidopsis, WUS orthologs in maize that specify the floral stem cell niche have not been functionally characterized. In maize, zag1 regulates FM determinacy with a lesser role in promoting carpel identity; currently, functional analyses have not been reported for the zea mays mads2 paralog (zmm2; Mena et al., 1996). zag1 and zmm2 expression domains overlap largely throughout the development of pistillate florets, where they mark the FM, as well as stamen and carpel primordia (Schmidt et al., 1993; Chuck et al., 2008). drl expression is excluded from meristems, but drl transcripts accumulate in lateral organs that initiate from meristems, including FM-derived carpel primordia (Fig. 3). Our findings using genetic interaction (Fig. 4A-E), GRN inference (Fig. 4H,I) and expression (Fig. 4J) analyses suggest that DRL1 and ZAG1 may auto-regulate and regulate each other, and potentially converge on a common set of downstream genes to control FM determinacy. We envision a scenario during floret development whereby DRL1 and ZAG1 are initially expressed independent of each other. Later, when their expression domains overlap, auto- and cross-regulation of each factor's expression is maintained and amplified; ultimately, DRL1 and ZAG1 synergistically regulate the expression of downstream genes (Fig. 4K).

Our results suggest that the drl genes interact differentially with the distinct developmental potentials of staminate UFMs, LFMs and pistillate UFMs (Figs 1 and 2). UFs and LFs differentially express key regulators (Cacharrón et al., 1999; Skibbe et al., 2008), potentiate differential effects of developmental regulators (Thompson et al., 2009), and derive from slightly different developmental trajectories of the SM (Irish, 1998). Perhaps akin to maize CLAVATA3/ESR-related (CLE) signaling peptides and the ZmFON2-LIKE CLE PROTEIN1 (FCP1)-FASCIATED EAR3 (FEA3) primordia-to-meristem feedback circuit (Je et al., 2016), a feedback signaling system from SM- and FM-derived lateral primordia involving the drl gene products could provide vital control of stem cell proliferation by integrating hormonal or metabolic cues from incipient and emerging primordia. With some 48 CLE genes currently reported in maize (Goad et al., 2017), it is tempting to speculate that differential interactions and/or regulation between drl, zag1 and CLE genes and/or gene products could provide non-cell-autonomous control of FM activity from lateral floral primordia. In support of this hypothesis, we found that DRL1 and ZAG1 are predicted to target the genes fcp1 and cle35-cle37 (Fig. 4I). Understanding the genetic relationship between CLE-encoding genes and the drl and zag1 genes in maize may contribute further to our understanding of factors that regulate FM activity.

Maize plants were grown in the field or the greenhouse. The drl1 and drl2 alleles used in this study were described previously (Strable et al., 2017). drl alleles were backcrossed to A619, B73, Mo17 and W22 inbred lines at least four times. The effects of drl1 alleles on floral development were fully penetrant in all backgrounds; backcrosses and F2 introgressions into B73 were used for analyses reported here. The ifa1 (B73, backcrossed four generations) allele was obtained from Sarah Hake (UC-Berkeley, CA, USA). The zag1-mum1 (B73, backcrossed many generations) allele was obtained from David Jackson (Cold Spring Harbor Laboratory, NY, USA) and na1-R (B73) was obtained from Phillip Becraft (Iowa State University, IA, USA). For quantitative phenotyping, sample sizes per genotype are indicated throughout the manuscript, along with mean±s.e.m. presented with significance calculated using two-tailed Student's t-tests. All experiments were performed with two or three independent biological replicates.

Allele tests and higher-order mutants were generated using the drl1-R, ifa1 and drl2-M alleles and the zag1-mum1 and na1-R alleles. The F₁ progeny from these crosses were grown to maturity and, in the case of triple mutant analysis, self-pollinated. The F₂ progeny were grown to maturity and screened for the drl1-R and drl2-M alleles (Strable et al., 2017) or for the zag1-mum1 allele by genotype (9242: AGAGAAGCCAACGCCAWCGCCTCYATTTCGTC; zag1_F2: GGAATCTGCTAGGCTGAGGC; and zag1_R2: GGTCGTTGAAGTCTTTCCGG). Genotyping primers and assays for ifa1, drl1-R and drl2-M, as well as DNA isolation and PCR conditions were described previously (Strable et al., 2017). Higher-order mutants between drl1-R, drl2-M and na1-R were screened by phenotyping.

Toluidine Blue O (TBO) (Sigma) staining was performed on mature spikelets. Briefly, TBO was dissolved in 1% sodium borate (w/v) to make a 1% stock solution (w/v). A 0.5% TBO staining solution was made immediately before use by diluting the stock solution with 1% sodium borate. Microtome sections of 10 μm, adhered to a microscope slide, were deparaffinized in Histo-Clear (National Diagnostics) (twice, 10 min each). Slides were passed through a graded ethanol series toward hydration, 1 min each (100%, 100%, 95%, 95%, 70%, 50%, distilled water) and stained in 0.5% TBO staining solution for 3 min. Slides were then passed through a graded series toward dehydration, 30 s each (50%, 70%, 95%, 95%, 100%, 100%) and Histo-Clear (three times, 5 min each). Slides were coverslip mounted with Permount (Fisher).

Field-grown ears 10 mm in length were fixed with 2% paraformaldehyde and 2% glutaraldehyde in cacodylate buffer (0.1 M) at pH 7.2 for at least 24 h at 4°C. After fixation, samples were rinsed three times (15 min each) in cacodylate buffer (0.1 M). Samples were then post-fixed in 1% osmium tetroxide in cacodylate buffer (0.1 M) for 1 h. After several washes with deionized water, samples were dehydrated through a graded ethanol series (25%, 50%, 70%, 85%, 95%, 100%), two changes each for 15 min. Samples were critical point dried using a Denton Vacuum Drying Apparatus, Model DCP-1. Dried materials were mounted on aluminium stubs with double-sided tape and colloidal silver paint and sputter coated with gold-palladium with a Denton Desk II Sputter Coater. Images were captured using a JEOL JSM-5800LV scanning electron microscope at 10 kV (Japan Electronic Optics Laboratory).

Field-grown 10 mm maize ears were fixed overnight at 4°C in formalin-acetic acid-alcohol (FAA). Samples were dehydrated through a graded ethanol series (50%, 70%, 85%, 95%, 100%) each 1 h, with two changes in 100% ethanol. Samples were then passed through a graded Histo-Clear (National Diagnostics) series (3:1, 1:1, 1:3; ethanol: Histo-Clear) with three changes in 100% Histo-Clear; all changes were 1 h each. Samples were then embedded in Paraplast Plus (McCormick Scientific), sectioned, and hybridized as described previously (Strable et al., 2017). Hybridizations were performed using antisense digoxygenin-labeled RNA probes: drl1 (Strable et al., 2017), drl2 (Strable et al., 2017), kn1 (Jackson et al., 1994), si1 (Bartlett et al., 2015), zag1 (Bartlett et al., 2015) and zyb15/yab8 (JS137-CGATCTCTACGCCGCAGC and JS138-GCAGACATACGCAAACATGGG).

Field-grown maize ears less than 8 mm long were dissected away from husk and prophyll primordia and placed individually in 100 µl Trizol (Thermo-Fisher) and stored at −80°C in a 1.5 ml Eppendorf tube until processing. To process, 400 µl Trizol was added and ear tissue was thawed and ground in the presence of Trizol using a plastic drill mount pestle. Total RNA was extracted as per the Trizol manufacturer and treated with RQ1 DNase (Promega) following the protocol outlined by the manufacturer, and converted to cDNA using RNA to cDNA EcoDry Premix (Double Primed) reagents (Takara Bio USA). The cDNA was diluted 1:1 with water, and 1.0 µl was used for PCR. Queried genes and primers used were: zag1 (zag1_F1 AGACAGCGAACATGATGGGG and zag1_R1 GACATAGTTGGTGCCAAGCC), si1 (si1_F1 CGAGGCGTACAAGAACCTGC and si1_R1 CAGTACCTCGGTTGCATTGC) and ubi1 (ubi_F1 TAAGCTGCCGATGTGCCTGCGTCG and ubi_R1 CTGAAAGACAGAACATAATGAGCACAGGC). PCR followed standard conditions using GoTaq Green Master Mix (Promega Corporation), Ta=58°C, 1 min. extension at 72°C for 33 cycles.

Publicly available transcriptome and proteome datasets that represent an atlas of tissues and developmental stages (Walley et al., 2016) were utilized to understand mRNA, non-modified protein and phosphoprotein quantities for drl1, drl2 and zag1 genes. Transcript abundance [in fragments per kilobase of transcript per million mapped reads (FPKM)], non-modified protein [in distributed normalized spectral abundance factor (dNSAF)] and phosphoprotein (in spectral counts) levels were retrieved directly from this public resource (Walley et al., 2016). We mined each of the seven gene regulatory networks that were generated and reported by Walley and co-workers (2016; Table S10) for regulator-target predictions by classifying DRL1 or ZAG1 as TF regulators and retrieving the set inferred regulator-target mRNA pairs with high-confidence edge scores. We report on high-confidence predicted target mRNAs in the YABBY, MADS-box and CLE gene families as well as other maize floral-expressed genes.

Genes referred to in this study include: drl1, GRMZM2G088309; drl2, GRMZM2G102218; kn1, GRMZM2G017087; si1, GRMZM2G139073; ubi1, GRMZM2G409726; zag1, GRMZM2G052890; zyb15/yab8, GRMZM2G529859.

We thank Harry Horner and Tracey Stewart at the Iowa State University Bessey Microscopy Facility for assistance with scanning electron microscopy and Pete Lelonek for plant care. Many thanks to Clint Whipple for generously sharing the si1 probe for RNA in situ hybridization and to Beth Thompson for discussions on ifa1. We thank Justin Walley for helpful discussions on incorporating GRN analysis. We are grateful for the many former undergraduate students, especially Sarah Briggs, Emery Peyton and Charlie Beeler, for their help in our summer genetics nurseries. Many thanks to Erin Irish and Erica Unger-Wallace for insightful discussions and comments on the manuscript. We also appreciate helpful suggestions and comments from three anonymous reviewers.