Floral meristem initiation and meristem cell fate are regulated by the maize AP2 genes ids1 and sid1

Meristems provide cells for all the tissues and organs of the plant, and understanding their activity is essential to decipher the evolution of plant architecture (Sussex and Kerk,2001). Meristems can be classified as determinate or indeterminate based on whether they terminate in the production of primordia. Inflorescence meristems (IMs), such as those of racemes, are usually indeterminate,producing an open number of lateral meristems. Floral meristems (FMs), by contrast, are usually determinate and terminate in the production of floral organs. The switch from indeterminacy to determinacy may also be considered in the light of developmental timing, or heterochrony. For example, if a switch to determinacy is delayed, the meristem will initiate extra lateral organs before terminating.

Inflorescence architecture in monocots is dependent upon the initiation of specific meristem types (McSteen et al.,2000). In maize, several lateral meristems with defined branching activity are initiated before FMs are made. These include branch meristems(BMs), spikelet pair meristems (SPMs) and spikelet meristems (SMs). Spikelets,the fundamental floral unit of all monocots(Clifford, 1987), consist of two bract leaves, called glumes, that enclose a variable number of florets. Maize SPM and SM are determinate. The SPM produces one SM and then terminates its activity by differentiating into an SM. Similarly, the SM produces one FM and then terminates by becoming an FM.

Several maize and rice mutants have provided clues as to how meristem fates are acquired. SM identity is controlled by the action of the orthologous genes branched silkless1 (bd1) in maize(Chuck et al., 2002) and frizzy panicle1 (fzp1) in rice(Komatsu et al., 2003). In bd1 and fzp1 mutants, the spikelet meristem is replaced by an indeterminate branch meristem. The bd1 and fzp1 genes encode members of the ERF family of APETALA2 (AP2) transcription factors, and function to repress indeterminate lateral branch meristem fates. SM determinacy is controlled by another AP2 transcription factor, encoded by the indeterminate spikelet1 (ids1) gene(Chuck et al., 1998). ids1 mutant spikelets are indeterminate and initiate extra florets instead of two. In rice, a similar phenotype is seen in the supernumerary bract (snb) mutant, which delays the floral transition and produces extra glumes before initiating florets. snb encodes an AP2 transcription factor similar to ids1, and might represent a paralogous gene (Lee et al.,2006). ids1 is targeted by the MIR172 microRNA encoded by the tasselseed4 (ts4) gene(Chuck et al., 2007). Mutations within the microRNA binding site of ids1 give rise to the dominant Tasselseed6 (Ts6) allele, which affects spikelet meristem branching and sex determination of floral organs(Chuck et al., 2007). Double mutant analysis has shown that ids1 mutants suppress the ability of the ts4 SM to initiate ectopic FMs. Thus, ids1 has diverse functions, affecting sex determination as well as meristem determinacy.

The orthologous floral homeotic genes LEAFY (LFY) of Arabidopsis and FLORICAULA (FLO) of Antirrhinum play major roles in the specification of FM identity in these diverse species (Weigel et al.,1992; Coen et al.,1990). Flowers on lfy mutants have characteristics of both floral and inflorescence meristems(Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992), whereas flowers on flo mutants are complete transformations to inflorescences. LFY activates expression of the MADS-box genes APETALA3 (Lamb et al.,2002) and AGAMOUS(Hong et al., 2003), both of which function in specifying floral organ identity. Mutations in the maize LFY/FLO orthologs, zfl1 and zfl2, affect floral organs and tassel branch number (Bomblies et al., 2003). Knock-down of the rice LFY/FLO ortholog, RFL, affects flowering time and panicle branching, yet spikelets and florets are still made (Rao et al.,2008).

Here we describe a paralog of ids1, sister of indeterminate spikelet1 (sid1), in maize. sid1 functions together with ids1 to specify the fate of several lateral meristems in the inflorescence. Fewer branches form in the tassel and fewer rows of spikelet pairs form in the ear. Both ear and tassel spikelet meristems are indeterminate and produce supernumerary organs. The tassel spikelets reiteratively produce bracts and never produce sex organs. The spikelets in the ear also produce many bracts, but eventually terminate in an ovule-like structure. Late-forming bracts in the ear are feminized owing to the ectopic expression of AGAMOUS-like genes in maize. Thus, ids1 and sid1 are needed to promote determinacy and produce sex organs,similar to LFY/FLO function in Arabidopsis and Antirrhinum.

Phylogenetic analysis was performed using MrBayes v. 3.1.2(Ronquist and Huelsenbeck,2003) Markov chain Monte Carlo (MCMC) algorithm using the Time-Reversible model (GTR) with a proportion of invariable sites (I) and a gamma distribution parameter (G). The analysis was run for 10,000 generations,sampling trees every 10 generations; 250 trees were discarded as burn in. Tree shows average branch lengths and posterior probabilities for all resolved nodes.

Poly(A)⁺ RNA gel blots and in situ hybridization were performed as described (Chuck et al.,2007). For sid1 in situ hybridizations, a 3′fragment outside the AP2 domain and including the 3′ UTR (bp 1301-1730)was used. This same fragment was used to probe northern blots. Probes for kn1 (Jackson et al.,1994), bd1 (Chuck et al., 2002) and zag1(Schmidt et al., 1993) were used as previously described.

Poly(A)⁺ RNA (200 ng) was isolated from 0.5 cm ear primordia from B73 and ts4-TP and used directly for ligation to the GeneRacer 5′-RACE oligo (Invitrogen) and reverse transcribed. One microliter of the reverse transcription reaction was used for RT-PCR using the GeneRacer 5′-RACE oligo in combination with the SID1 53R oligo(5′-AACCATACCCAACAGTGGCGACTG-3′) located 280 bases from the microRNA binding site. The wild-type cleavage product was cloned into pGEM-T Easy (Promega) and sequenced using M13 primers.

Double mutants between ids1-mum1 and sid1-mum4, between ids1-Burr and sid1-mum3, and between ids1-mum1,sid1-mum1 and ts4-A were made by selfing heterozygotes. Homozygotes for sid1 were scored by PCR using sid1 oligos 607 and 608 (5′-ATCAGCGTGTGCCTAGCATTTCTTCCCT-3′ and 5′-CAGTCCCTGCACAATTGGACGACACA-3′). ids1 homozygotes were scored using ids1 oligos 804 and 674(5′-ATCGCAGCTCGATCGTATG-3′ and 5′-TACAGGGGCGTCACCTTCTA-3′). ts4-A homozygotes were scored using ts4 oligos 7F and 7R(Chuck et al., 2007).

Previous work identified five potential MIR172 targets(Chuck et al., 2007). AY109248.1, now referred to as sid1, had high sequence similarity to ids1, especially within the AP2 domain, where it displayed 96% amino acid identity (Fig. 1A). A Bayesian phylogenetic tree of MIR172 targets in rice and maize showed that sid1 shares a common ancestor with ids1(Fig. 1B), and is likely to be orthologous to the rice snb gene, which controls the SM-to-FM transition (Lee et al., 2006). To determine the function of sid1, Mutator (Mu) transposons were employed to generate loss-of-function alleles(Bensen et al., 1995). Four Mu insertions were isolated in sid1, three in exons and one in an intron (Fig. 1C). sid1-mum1 and sid1-mum2 alleles had no transcript, whereas the sid1-mum3 and sid1-mum4 alleles had longer transcripts,possibly owing to alternative splicing of the Mu element(Fig. 1F).

A combination of RT-PCR and RNA gel blots was performed to determine sid1 transcript levels in the wild type and in ts4 and ids1 mutants. RT-PCR demonstrated that sid1 is widely expressed in all tissues assayed, including roots, leaves and inflorescences,and is elevated in tassels from ts4 mutants(Fig. 1D). An increase in sid1 transcript levels was also detected by gel blots with RNA from ts4 mutant tassels (Fig. 1F) and ears (Fig. 1G), indicating that the sid1 transcript might be subject to negative regulation by ts4. In support of this, RNA cleavage assays of sid1 transcripts from ts4-TP mutants, as compared with those from wild type, showed differences in the amount of cleavage within the microRNA binding site. In ts4-TP mutants, most transcripts were cleaved 3′ of the microRNA binding site(Fig. 1E). By contrast, in wild-type RNA, almost all transcripts were cleaved between bp 10 and 11 of the microRNA binding site, as expected. Some cleavage of sid1 still occurred in ts4-TP within the microRNA binding site, suggesting that sid1 is also targeted by other members of the MIR172 family,of which there are at least four (Chuck et al., 2007). sid1 expression was also increased in ids1 mutant ears as compared with wild type(Fig. 1G). This finding indicates that sid1 might be under negative regulation by ids1. AP2 genes are predicted to be negatively autoregulated in both Arabidopsis (Schwab et al.,2005) and wheat (Simons et al., 2006). Compensation between redundant genes has been documented in numerous cases, for example with the PIN genes in Arabidopsis. Mutations in pin7 result in ectopic expression of PIN4, and PIN1 is ectopically induced in pin2mutants (Vieten et al.,2005).

Maize is a monoecious plant, in which male and female flowers are borne on distinct inflorescences. Sex determination occurs through abortion of carpel primordia in the tassel and arrest of stamen primordia in the ear(Cheng et al., 1983). Ears and tassels also differ by the presence of BMs, which form side branches in the tassel and are absent from the ear (Fig. 2A, Fig. 3A). SPMs initiate in ordered rows in a spiral phyllotaxy from the inflorescence meristem and in a distichous phyllotaxy along the tassel branches(Fig. 3A,B). Each SPM initiates an SM and converts to a second SM. Each SM initiates two sterile leaves called glumes, followed by two lemmas, each containing FMs in their axils. The FM initiates a palea, lodicules, stamens and finally a pistil, or silk(Fig. 3C,D). In tassels, the pistils abort to produce a pair of staminate florets(Fig. 2H), whereas in the ear,the lower floret and the stamens abort to produce a single pistillate floret(Fig. 2D)(Cheng et al., 1983).

Since sid1 mutants appeared to be phenotypically normal (data not shown), double mutants with ids1 were analyzed. These double mutants affected all the lateral meristems of the inflorescence. The tassels of the ids1-Burr;sid1-mum3 double mutant(Fig. 2A) had fewer branches than with ids1-Burr alone(Kaplinsky and Freeling,2003). Normal female inflorescences initiate seven to eight rows of SPMs, but ids1-Burr;sid1-mum3 double mutants initiated a maximum of four rows, and often had bare rachis in place of the missing rows(Fig. 2B). Although the ears of the double mutant appeared to initiate silks, seed set was reduced 5- to 8-fold compared with ids1-Burr(Fig. 2C), indicating that floral organ function was affected. In fact, most of the seeds that formed failed to germinate. ids1-Burr ear spikelets initiate extra lateral florets in the axils of lemmas (Fig. 2E). In the double mutant, these extra lateral florets were absent, and the spikelet terminated in a floret-like structure(Fig. 2F,G). This terminal structure was enclosed by several bracts that have silk-like characteristics at their tips (Fig. 2G). ids1-Burr tassel spikelets also initiate several functional florets(Fig. 2I). The tassels of the double mutant, however, did not initiate any florets, and continuously initiated bracts instead (Fig. 2J). A double mutant between different alleles, ids1-mum1and sid1-mum4, appeared identical to ids1-Burr;sid1-mum3(data not shown), and thus these phenotypes are not allele-specific.

Although ids1 is targeted by ts4, ids1 mutants do not completely suppress the ts4 mutant phenotype because the ear of the double mutant still resembles that of ts4 single mutants(Fig. 2K)(Chuck et al., 2007). Thus, at least one other target gene must be responsible for the continued presence of the ts4 ear phenotype in the ids1;ts4 double mutant. Double mutants between sid1-mum1 and ts4-A resembled ts4-Aalone in both the tassel (Fig. 2L) and ear (Fig. 2N, left). However, when these double mutants were also made heterozygous for ids1-mum1, the tassel was normalized and resembled a weak ts4 mutant phenotype (Fig. 2M), whereas the ear resembled that of wild type(Fig. 2N, right). This result supports the RNA analysis (Fig. 1D,F,G) and indicates that sid1 is a ts4 target gene, although its suppressive effects can only be observed in combination with ids1. The ids1-mum1;sid1-mum1;ts4-A triple mutant resembled the ids1-Burr;sid1-mum3 double mutant and completely suppressed the ts4 sex-determination phenotype in the tassel(Fig. 2A, right) and extra branching phenotypes in the ear (data not shown). Thus, ids1 and sid1 together are epistatic to ts4.

Scanning electron microscopy was used to examine the ids1;sid1double mutant phenotype more closely. ids1-mum1 ear tips normally initiate six rows of spikelet pair meristems(Fig. 3E). ids1-mum1ear spikelets normally initiate several extra lemmas, each of which contains FMs in their axils (Fig. 3F). These extra FMs are fully functional, and initiate floral organs in a normal pattern (Fig. 3G). By contrast, ids1-Burr;sid1-mum3 tassels initiated fewer BMs(Fig. 3H), and only four rows of SPM (Fig. 3I). The SM of the tassel in the double mutant was indeterminate and continuously initiated lemma-like bracts with no FMs (Fig. 3J). The ear of the double mutant also initiated fewer rows(Fig. 3K, Fig. 4G). In contrast to the tassel, the SM of the ear terminated in an ovule-like structure after initiating several bracts (Fig. 3L). No stamens or lodicules were observed, indicating that the normal pattern of floral organ initiation is not followed in the double mutant. Later in development, the tips of the bracts began to differentiate structures that resemble silks (Fig. 3M).

To examine the expression of sid1 more precisely, in situ hybridization on wild-type and mutant tissue was performed. sid1transcript was present in both the SPM and SM of wild type(Fig. 4A), but was absent from the sid1-mum1 inflorescence (Fig. 4B). After florets were initiated, sid1 transcript was found in all floral organs, including the carpels and stamens, as well as the lower FM (Fig. 4C). In ids1-mum1 florets, a similar pattern was seen(Fig. 4D), although expression was expanded owing to the initiation of extra meristems. The ts4 ear tip showed ectopic sid1 expression near the inflorescence meristem tip, as well as higher-level expression in the lateral meristems(Fig. 4E).

In situ hybridization using meristem-specific markers was carried out to further understand the basis for the indeterminacy and organ identity defects in ids1;sid1 double mutants. The maize knotted1(kn1) gene is expressed in the indeterminate cells of the meristem and is downregulated upon organ initiation(Jackson et al., 1994). Cross-sections of ids1-Burr ears compared with ids1-Burr;sid1-mum3 ears showed that the number of rows of SPMs was reduced (Fig. 4, compare F with G). kn1 expression in the spikelets of the double mutant persisted after the initiation of several sterile bracts(Fig. 4H), indicating that the meristem is indeterminate. The branched silkless1 (bd1) gene is expressed in a semi-circular domain at the base of the SM and disappears upon floret initiation in maize and rice(Chuck et al., 2002). In young ids1-Burr;sid1-mum3 spikelets, the bd1 expression pattern appeared normal (Fig. 4I). Later in development, however, instead of disappearing, this expression pattern persisted (Fig. 4J),demonstrating that the meristem had retained SM identity longer than normal.

The maize AGAMOUS-like MADS-box genes zmm2 and zag1 are duplicate genes that mark the FM as well as stamens and carpels (Schmidt et al., 1993; Mena et al., 1996)(Fig. 4K,M). They are not expressed in leaf-like organs, such as palea or lemma. zag1 mutations cause extra carpels to form that fail to fuse into a functional silk(Mena et al., 1996). At early stages of spikelet development, we saw no expression of zmm2 (data not shown) or zag1 (Fig. 4N, inset) in the ids1;sid1 double mutant. However,ectopic zmm2 expression was observed later in the SM of the double mutant as well as in the carpelloid bracts near the apex(Fig. 4L). zag1 was ectopically expressed in these same regions(Fig. 4N). The ectopic expression of AGAMOUS-like MADS-box genes in the ids1;sid1double mutant is consistent with the role of the Arabidopsis AP2 gene as a negative regulator of AGAMOUS(Drews et al., 1991). Finally,the maize AP3 ortholog, silky1, which is necessary for stamen and lodicule patterning (Ambrose et al., 2000), was not expressed in the double mutant at early (not shown) or late (Fig. 4P)stages, whereas it was detected in the stamens or lodicules of wild-type controls (Fig. 4O).

Based on previously described differences between the ts4 and Ts6 mutant phenotypes (Chuck et al., 2007), it was postulated that at least one other gene was targeted by the ts4 microRNA in addition to ids1. sid1appears to be that gene based on four criteria: its expression levels are elevated in ts4 mutants (Fig. 1D,F,G); its cleavage is reduced in a ts4 mutant background (Fig. 1E); it is ectopically expressed in ts4 mutant inflorescences(Fig. 4E); and sid1mutants enhance suppression of the ts4 mutant phenotype by ids1-mum1 (Fig. 2N). Previous work showed that MIR172 represses its AP2 target genes at the level of translation (Chen,2004; Aukerman and Sakai,2003). Other groups have found that MIR172 repression may act at both the level of translation and transcript cleavage(Schwab et al., 2005). We showed that IDS1 is regulated at the level of translation by ts4(Chuck et al., 2007). Our analysis of sid1 suggests that it is regulated at the level of transcript stability, although we cannot rule out an additional level of translational regulation. Thus, it is likely that ts4 regulates AP2 transcripts using both modes of regulation.

Lack of negative regulation of ids1 in Ts6 mutants leads to SMs that initiate extra florets in a random phyllotaxy(Irish, 1997; Chuck et al., 2007). Surprisingly, ids1 mutants also initiate extra florets, but in a defined distichous phyllotaxy. Thus, the loss-of-function and gain-of-function phenotypes for ids1 seem similar with regard to floret initiation. This conflicting result can be explained by the function of its redundant paralog, sid1, which masks the ids1 loss-of-function null phenotype. Loss of both sid1 and ids1 leads to loss of FM initiation in the tassel and ear. Therefore, the gain-of-function phenotype of ids1 (Ts6) shows that it is sufficient for FM initiation,and the complete loss of function of ids1 and sid1 shows that it is necessary for FM initiation.

Loss of ids1 and sid1 also results in a reduction in the number of lateral meristems that initiate from the inflorescence meristem. This result seems to indicate a role for ids1 and sid1 in maintaining the stem cell niche, much like AP2 genes in Arabidopsis (Aida et al.,2004; Wurschum et al.,2006). The use of microRNA-resistant forms of AP2demonstrated that this property acts through the WUSCHEL(WUS) pathway, which plays a key role in stem cell maintenance(Zhao et al., 2007). However,a simple comparison of the inflorescence meristems of the wild type and ids1;sid1 mutant (Fig. 4A,H) showed that they were approximately the same size. This indicates that the more likely function of ids1 and sid1 is to regulate initiation of lateral meristems in the inflorescence, rather than regulating stem cell maintenance.

The carpelloid bract defects in ids1-Burr;sid1-mum3 ear spikelets appear to be caused by ectopic expression of AGAMOUS-like MADS-box genes in the lateral organs of the spikelet. AP2 mutants in Arabidopsis display carpelloid sepals owing to ectopic expression of AGAMOUS in those organs (Drews et al., 1991). The ids1;sid1 floral organ phenotype demonstrates that this property is conserved in maize. Interestingly, the ectopic in situ expression of zag1 and zmm2 was not observed in the tassel (data not shown), which might indicate that these genes have divergent functions in the male and female inflorescences, as previously suggested (Mena et al.,1996).

The ids1;sid1 tassel does not initiate FMs or floral organs. This function is unique to maize, because ap2 mutants in Arabidopsis continue to make flowers. A role for ap2 in FM fate is seen in double mutants with ap1. Arabidopsis ap2;ap1 mutant flowers are less determinate, and initiate axillary structures in a helical pattern that develop as pistils with occasional stamens(Irish and Sussex, 1990; Shannon and Meeks-Wagner,1993). The lfy mutant of Arabidopsis displays striking similarities to maize ids1;sid1. Both lfy and ids1;sid1 mutants are indeterminate, initiate several bracts before terminating in carpel-like structures(Huala and Sussex, 1992), and lack expression of B-function genes such as APETALA3/silky1(Weigel and Meyerowitz, 1993). The lfy phenotype is most obvious on early-formed flowers and, like ap2, is enhanced in double mutants with ap1(Huala and Sussex, 1992; Weigel et al., 1992; Shannon and Meeks-Wagner,1993). Thus, LFY, AP1 and AP2 are all thought to play a role in the transition from IM to FM in Arabidopsis.

The maize lfy double mutant, zfl1;zfl2, displays additional similarities to ids1;sid1 in that fewer tassel branches form and stamen development is affected(Bomblies et al., 2003). Unlike ids1;sid1, however, zfl1;zfl2 mutants display indeterminacy in the carpel and FM, and not in the SM. Moreover, the ids1;sid1phenotype is distinct in that both FMs and sex organs fail to form. The uniqueness of the ids1;sid1 phenotype might be due to its effects on the fate of the SM, which normally initiates sterile bracts before florets are made. If ids1 and sid1 are required for the SM to transition to FM initiation, then a block at this step could lead to the formation of multiple sterile organs. Since floral bracts are present in both Arabidopsis and maize, these pathways are likely to be parallel in both monocots and dicots. One intriguing possibility is that the divergence of monocots and dicots has led the AP2 genes to substitute for LFY function in the grasses. Analysis of sid1 and ids1 in other genera closer to the monocot/dicot divide will elucidate the evolution of AP2 versus LFY genes in FM fate.

We thank D. Hantz for greenhouse maintenance; B. Thompson, N. Bolduc, C. Lunde and H. Candela for helpful comments on the manuscript; and Devin O'Conner for phylogenetic analysis. This work was supported by National Science Foundation (NSF) grant DBI-0604923 and by USDA-ARS Current Research Information System (CRIS) grant 5335-21000-018-00D to S.H., and by Cooperative State Research, Education and Extension Service (CSREES) grant 2004-35301-14507 to G.C.