Combinatorial control of meristem identity in maize inflorescences

Plant growth consists of reiterative developmental events that determine the morphology of the mature organism. Meristems, self organizing pools of undifferentiated cells, give rise to most plant organs including other meristems. The above ground portion of the plant is elaborated from the shoot apical meristem (SAM), which produces leaves, stems and axillary meristems(together called a phytomer) in a repetitive manner(Galinat, 1959;Steeves and Sussex, 1989). If meristem maintenance is disrupted, production of phytomers ceases and plant development arrests (Long et al.,1996; Vollbrecht et al.,2000). Therefore, proper plant development depends on normal meristematic activity. The reproductive structure of plants, the inflorescence, is produced by an inflorescence meristem (IM), which is derived from the SAM and/or axillary meristems. Inflorescences exhibit a wide range of meristematic behaviors in different plants, leading to the varying reproductive morphologies seen in the plant kingdom. Secondary and higher order meristems can be initiated from the primary IM. Each order of meristems represents an order of branching, thus meristem function and inflorescence shape are directly related (Kellogg,2000). In the grasses, inflorescences range from simple and unbranched spikes to highly branched panicles. This variability can be observed within single species (Chapman and Peat, 1992). During the domestication of corn, Zea mays,from teosinte, there has been an increase in inflorescence complexity that reflects human selection for agronomically important traits(Doebley and Stec, 1991;Galinat, 1996;Iltis, 1988).

Maize is a monoecious species that has two types of inflorescences, the male tassel at the apex of the plant, and female ears in the axils of leaves. Tassel and ear development are similar until after flowers are initiated, at which point organ abortion in the tassel and the ear produce separate unisexual inflorescences (Dellaporta and Calderon-Urrea, 1994; Irish,1996). At the top of the plant the SAM elongates and converts into an IM, which gives rise to the tassel. At approximately the same time, the axillary meristems initiate a prophyll (the first leaf on a shoot) and eight to 14 husk leaves, and then convert into a lateral IM, which produces the ear. The top (youngest) ear primordium grows faster than lower primordia and often develops into a single dominant ear. The IM gives rise to 2°, 3°, and 4° meristems (Fig. 1A). Each IM produces an indeterminate number of spikelet pair meristems (SPM,2°) in an acropetal (from the base to the apex), polystichous (multiple rows) manner. In the tassel, the first few SPMs become branch meristems and produce long tassel branches that initiate SPMs in a distichous pattern. All other SPMs initiate one spikelet meristem (SM, 3°) and then convert into SMs. Each SM then initiates a pair of bract-like glumes. Subsequently, each SM initiates the lower floret meristem (FM, 4°) and converts into the upper FM. The FMs make the floral organs, the palea/lemma, lodicules, anthers and pistils. The pistils abort in the tassel, while the anthers and lower pistil abort in the ear (Cheng et al.,1983). This results in a single kernel developing from each ear spikelet. A large number of mutants that affect these meristem initiation and identity conversion events have been identified and make the maize inflorescence a powerful system for studying meristem function(McSteen et al., 2000;Veit et al., 1993).

Maize inflorescence mutants affect tassel and ear development in many ways. Interestingly, mutant phenotypes often differentially affect the two inflorescence types (Postlethwait and Nelson, 1964). These mutants interact with each other, often synergistically. The class II tasselseed mutants (ts4, Ts6) affect both floral organ abortion and branching. Pistil abortion in the tassel is inhibited resulting in a feminized tassel with seeds. In addition, SPMs fail to convert into SMs in ts4, and SM to FM transitions are delayed inTs6. Both of these mutations result in extra branching in the ear(Irish, 1997a;Irish, 1997b). The ramosa mutants (ra1, ra2, ra3) have indeterminate SPMs; these make long branches that produce many SMs(Postlethwait and Nelson,1964; Veit et al.,1993). indeterminate spikelet1 (ids1) mutants delay the SM to FM conversion resulting in the production of extra florets,which is a more branched and less determinate condition(Chuck et al., 1998).indeterminate floral apex1 (ifa1) affects determinacy of all inflorescence meristems, resulting in extra SPMs, SMs and an indeterminate FM.ids1 and ifa1 interact synergistically, with SMs converting into branch meristems in the ear and into SPMs in the tassel(Laudencia-Chingcuanco and Hake,2002).

Perturbations of regular rowing and kernel orientation are sensitive indicators of defects in inflorescence development. Increased or decreased numbers of florets produce an easily recognizable reversed germ orientation(rgo) phenotype (Fig. 2D). Several reversed germ and fused kernel mutants have been previously described in maize. Kiesselbach describes the occurrence of reversed and fused kernels in various inbreds and hybrids, as well as documenting reports of these phenotypes dating back to 1810(Kiesselbach, 1926). In this and subsequent reports (Jackson,1996b; Joachim,1956), the reversed orientation of the kernels was attributed to the development of the lower instead of the upper floret. Fused kernels were attributed to the development of both florets in a spikelet. The development of both florets occurs in Country Gentleman sweet corn and results in reversed kernels (Micu et al., 1983). Diverse patterns of floret development are found in the grasses; basal fertile florets with non-fertile apical florets are found in the pooids, while the opposite pattern occurs in the panicoids(Chapman and Peat, 1992).

The reversed germ orientation1 (rgo1) mutant was isolated by Sachan and Sarkar. They suggested that the phenotype was due to the differential growth of the integuments and the nucellus causing rotation of the kernel (Sachan and Sarkar,1980). In this paper, we show that the rgo1 reversed kernel phenotype is actually due to the production of an extra floret by the spikelet meristem. We demonstrate that rgo1 and ids1 have overlapping functions in regulating meristem identity. The rgo1; ids1double mutant is synergistic and affects more meristem types than either mutant alone. A dosage based model is proposed for the regulation of meristem identity conversion in the maize inflorescence.

A rgo1-Sarkar stock (904G) was obtained from the Maize Genetics Cooperation Stock Center (MGCSC, Urbana, IL)(Jackson, 1996a). This was introgressed seven times into B73 (Pioneer Hi-Bred, Des Moines, IA). The B73 introgression was used for complementation and allelism crosses and double mutant analyses. Plants were grown in fields in Berkeley, CA and San Jose, CA during the summer. Plants were grown in greenhouses in Berkeley, CA and on Molokai, HI during winter seasons. The phenotype of rgo1 did not vary in these different growth conditions. ids1-mum1(Chuck et al., 1998) was a gift from G. Chuck (USDA, ARS, Albany, CA). ids1-VI and ids1-Burrwere gifts from J. Jackson (MGCSC, Urbana, IL). Originally described asrgo2-VI and rgo2-Burr, complementation and allelism tests with both ids1-mum1 and rgo1, as well as mapping data,showed that they are alleles of ids1(Jackson and Kaplinsky, 2000). To observe double mutants, plants used for the ids1; rgo1complementation test were selfed. The genotype of representative double mutant individuals was confirmed molecularly using an ids1 cDNA and umc114 as RFLP probes, and umc1267 primers to detect a rgo1-linked SSR. Some double mutants were constructed by crossing wx1 rgo1 lines generated during mapping (see below) to ids1 plants. Selfed progeny from this cross segregated wx1 kernels, which were enriched for rgo1.Genetic nomenclature follows the standards for maize genetics(Burr et al., 1995); unlinked genes are separated by a semicolon, linked genes by a space.

Wx1 rgo1 plants were crossed to a wx1 stock and 33wx1 translocation lines representing 18 chromosome arms. The progeny of these crosses were self-pollinated and the resulting waxy (wx1)and opaque (Wx1) seed were planted separately. Linkage to an arm is indicated by repulsion from a given translocation marked by wx1(Laughnan and Gabay-Laughnan,1996). After mapping to a chromosome arm, finer molecular mapping was performed using RFLP and SSR markers.

1-3 cm developing ears were dissected from 6- to 8-week old plants and fresh tissue was used for imaging. 1-3 cm tassel primordia were fixed in FAA(3.7% formaldehyde, 5% glacial acetic acid, 50% ethanol) overnight at 4°C,dehydrated through an ethanol series, and critical-point dried. Glumes were dissected away to expose the floret primordia, and the samples were then sputter coated with 15 nm of gold/palladium (Polaron SEM coating system). An Electroscan E3 Environmental Scanning Electron Microscope (ESEM) was used to image samples at 20 kV. Images were digitally captured using IAAS software(Electroscan Corp.). Electron microscopy was performed in the Electron Microscope Laboratory at the University of California at Berkeley.

Tissue was fixed overnight at 4°C in FAA, dehydrated through an ethanol series, and stained with 0.1% Safranin. The samples were then embedded in paraffin (Paraplast, Oxford Labware, St Louis, MO). The samples were mounted on stubs, and 10 μm sections were cut and then mounted on slides. The slides were deparaffinized in Histoclear (National Diagnostics, Atlanta, GA),hydrated through an ethanol series, and stained for 30 seconds to 1 minute with 0.05% Toluidine Blue O. Slides were then rinsed, dehydrated, and mounted with Merckoglas (Mikroskopic, Germany)(Ruzin, 1999). Images were digitally captured using a Zeiss Axiophot light microscope.

Tassel phenotypes were quantified by counting the number of tassel branches, the number of spikelets on the central spike of the tassel, and the number of florets per spikelet. Between 20 and 50 spikelets from the central spike were counted to determine the number of florets per spikelet, and the numbers from siblings were pooled.

To analyze the rgo1 phenotype, the rgo1 mutation was introgressed into a B73 background. During vegetative growth, rgo1plants are indistinguishable from wild-type siblings. However, rgo1ears have irregular rows of kernels compared to wild-type siblings(Fig. 2A,B). Closer examination of rgo1 ears reveals kernels with their embryos facing both the tip and the base of the ear (Fig. 2B,D). Often, reversed and normal orientation kernels are found in the same spikelet surrounded by a shared pair of glumes. rgo1 ears also contain fused kernels, each containing an embryo(Fig. 2B). The frequency of fused kernels varies from ear to ear. Open pollination of rgo1 plants produced fused kernels where one kernel was fertilized by a pollen grain containing a color gene, the other without.Fig. 2E and F show the top and bottom of a fused kernel, each with a visible embryo.

In wild-type female spikelets the upper floret develops into the kernel with the embryo facing the tip of the ear, while the lower floret arrests and aborts. In wild-type male spikelets, two florets develop and produce three anthers each (Fig. 2G). Inrgo1 tassels spikelets have three florets, each with three anthers,for a total of nine anthers per spikelet(Fig. 2H). The penetrance of the tassel phenotype is background dependent, and ranges from 39% in B73 to 91% in Gaspe Bay Flint (GBF). The total number of spikelets and number of tassel branches are not affected in rgo1 mutants(Table 1).

Self pollinations of heterozygous rgo1/+ plants produced 25%(n=48) mutant progeny, and crosses of rgo1/+ heterozygotes to rgo1 homozygotes produced 50.6% (n=81,χ ²=0.012, d.f.=1, P=0.91) mutant progeny. This indicates that rgo1 behaves as a single recessive mutation when introgressed into B73. waxy1 (wx1) marked reciprocal translocation stocks were used to map rgo1 to a chromosome arm.rgo1 was in repulsion to all of the wx1 marked translocations as well as to a wx1 stock, indicating thatrgo1 is located on chromosome 9 and linked to wx1. Test crosses place rgo1 and wx1 approximately 11 cM apart. Further mapping with RFLP and SSR molecular markers locate rgo1within 2 cM of the SSR marker umc1267 in bin 3, close to the centromere.

To define the developmental defects in rgo1 mutants, early ear development was observed using electron microscopy and histology. Inflorescence meristems (IMs) of rgo1 ears initiate multiple rows of SPMs in the normal polystichous pattern and appear identical to wild-type IMs(Fig. 3A,B). SPMs inrgo1 plants initiate a SM correctly from the flank of each SPM(Fig. 3C,D), and are thus not responsible for the disrupted rowing observed in rgo1 ears(Fig. 2B).

Differences between normal and mutant development become evident upon examination of the SM. Normally each SM initiates a set of outer and inner glumes, and then the lower floret meristem(Fig. 3E). The SM itself converts into the upper floral meristem. rgo1 SMs also initiate two glumes and a FM, but then continue to make a second FM at 180° from the first FM (Fig. 3F). Subsequently both wild-type and rgo1 SMs become FMs and initiate floral primordia (Fig. 3G,H). However, the orientation of the floral organs on rgo1 spikelets is at 180° from normal, with the middle anther primordium on the same side as the outer glume (Fig. 3H), as opposed to the inner glume as in wild-type spikelets(Fig. 3G). Later in development(further down the developing ear), the FM makes the gynoecial ridge, two fused carpels that elongate to become the silk(Nickerson, 1954). Like the anther primordia, the orientation of the developing silk on the top flower is reversed in rgo1 spikelets (Fig. 3I,J). In rgo1 spikelets the lowest FM arrests and aborts similar to wild-type development, but in many spikelets the second flower continues to develop (arrowheads in Fig. 3J). The developmental defects in tassel spikelets mirror those observed in ear spikelets, with no defects observed until FMs are initiated. Wild-type SMs initiate only one FM (Fig. 4A), while rgo1 SMs initiate two(Fig. 4B). The orientation of the third flower in rgo1 tassel spikelets is reversed compared to wild-type apical florets (Fig. 4C,D).

During normal development the lower flower aborts while the upper flower makes a silk (Fig. 5A). The flower is then fertilized, and develops into a kernel. The lower flower undergoes programmed cell death which is detectable using the TUNEL assay for DNA fragmentation (Jones et al.,2001) and the accumulation of autoflorescent compounds in the cell wall (data not shown). After cell death, collapsed cell walls are all that remains of the aborted flower (af in Fig. 5E). In rgo1 plants, developing flowers have several fates, reflecting the variable expressivity of the mutation. Mirroring wild-type development, the apical flower can develop while the two lower flowers abort (Fig. 5B,F). However, in many rgo1 spikelets the second flower does not abort and the two upper flowers develop while the lowest flower aborts(Fig. 5C,G). Two kernels are produced in spikelets where the two apical flowers develop. The top (third)flower produces a reversed kernel, while the middle (second) flower's kernel has a normal orientation (Fig. 1). These two flowers can fuse during development, producing fused kernels (Fig. 5D,H). In spikelets with two developing flowers, the flowers are crowded and displace each other resulting in the disrupted rows of kernels observed in maturergo1 ears (Fig. 2B).

indeterminate spikelet1 (ids1) is a maize gene with homology to the AP2 class of transcription factors. Mutations of theids1 gene produce more than the usual two florets per spikelet. This phenotype has been interpreted as having less determinate SMs. In some backgrounds the spikelet becomes totally indeterminate. ids1 is expressed in both SPMs and SMs, although the ids1 phenotype only affects spikelet meristems (Chuck et al.,1998). Because of the production of extra florets, ids1mutant ears also have a reversed germ and disturbed rowing phenotype(Fig. 6). The lack of a SPM phenotype despite ids1 mRNA expression suggests redundancy with another gene or genes, and given the similarity of the rgo1 andids1 phenotypes, we investigated whether the double mutant would display a SPM phenotype.

Homozygous rgo1 and ids1 mutants were crossed to each other. The resulting ids1/+; rgo1/+ plants were grown and self-pollinated to produce the double mutant. Since ids1 maps to chromosome 1 and rgo1 maps to chromosome 9, it was a surprise that the ears of the doubly heterozygous plants had disturbed rows and reversed kernels, reminiscent of the ids1 and rgo1 single mutants(Fig. 6). When two recessive mutations at different loci fail to complement each other, the genetic behavior is known as nonallelic noncomplementation(Yook et al., 2001). Penetrance of this phenotype was 26% in ids1-Burr, 84% inids1-mum1 and 100% with ids1-VI. Penetrant ids1-VIand ids1-mum1 ears were used for further studies. Nonallelic noncomplementation suggests that genetically, rgo1 and ids1have overlapping functions. The occurrence of a mutant phenotype with the loss of two functional copies (either a rgo1 or ids1 homozygote or the ids1/+; rgo1/+ double heterozygote) implies that at least three wild-type copies of these two genes are needed for correct regulation of spikelet meristem activity.

In order to observe double mutants, families of the selfed progeny of double heterozygote plants were grown. Three classes of F₂ ears were observed in these families: normal ears, ears with reversed kernels and disrupted rowing like rgo1 or ids1 single mutants, and ears with proliferated spikelets and long branches(Fig. 6). The expectation based on the observation of nonallelic noncomplementation is that 11/16 of the plants (all plants missing two or more copies of ids1 orrgo1) would exhibit a mutant phenotype, and 1/16 would be double mutants. Table 2 lists the observed numbers of ears in each of these phenotypic classes. All families segregated the three phenotypic classes as expected, although the severity of the double mutant phenotype varied from family to family.

While single mutants only affect the SM, the ids1; rgo1 phenotype affects the SM as well as the SPM. To analyze the double mutants, we define meristems that produce glumes or florets as SMs, and meristems that produce spikelets as SPMs. A range of effects on the SM is observed, from retardation of identity transitions to identity reversions. As expected from the single mutant phenotypes, extra florets are produced by each SM, resulting in reversed and fused kernels (Fig. 7A,E). In addition, the double mutant SMs make supernumerary glumes (Fig. 7B,G). The SM can initiate one or more florets, elongate into a branch, and then initiate one or more spikelets (Fig. 7C,E). Both the primary and reiterated spikelets are identified by the presence of subtending glumes (Fig. 7C,D). In cases where the SM is producing spikelets, it reverts to SPM identity.

SPMs produce too many spikelets in the double mutant. This phenotype can be restricted to the base of the ear (Fig. 6, Fig. 7A) where the SPM produces multiple SMs. In the most severe cases, all of the SPMs on the ear become indeterminate (Fig. 7F), and produce multiple spikelets(Fig. 7G). These spikelets make many glumes (arrowheads, 7G) and only rarely produce florets as evidenced by the absence of silks. Since glumes are produced by the SM, in these spikelets the transition from producing glume primordia to floret primordia is delayed.

Unlike the synergistic interaction observed in the ear, rgo1; ids1tassels have an additive phenotype. More florets are produced by each spikelet, but there is no evidence that the SMs in the tassel have changed identities or that SPMs are affected (Table 1).

The progression from vegetative to floral meristem identities during plant growth is critical for normal plant reproduction. We have shown that progression of meristem identity from the spikelet to the floret stage is retarded (i.e. the spikelet meristem does not convert into a floret meristem,and thus continues to initiate florets) in the rgo1 mutant(Fig. 8A,B). As a result, extra flowers are produced, and because of the distichous pattern of floret primordia initiation on the spikelet axis, kernels with a reversed germ orientation are observed (Fig. 1B). Many rgo1 ear spikelets form two fertile flowers. Some of these bi-floral spikelets have fused flowers(Fig. 5D). This may indicate that conversion of the SM to FM identity occurred before two distinct flowers formed. The abnormal survival of sub-apical florets may be a result of a lack of suppression by the apical flower(Novoplansky, 1996).

If ids1 is placed into the framework of meristem identity changes it would perform a similar role as rgo1(Fig. 7C), promoting the SM to FM change. The overlapping functions of rgo1 and ids1revealed by their mutants and genetic interactions support the idea that these genes perform similar biological functions. In some double mutant spikelets,the SM reiterates more SMs (Fig. 7D,E) or makes supernumerary glumes(Fig. 7G) before initiating any FMs. The phenotype indicates that SM identity changes are inhibited or even reversed. This is consistent with a meristem identity role for bothrgo1 and ids1. In addition, in the double mutant the SPM also produces long branches with multiple SMs(Fig. 7A,F). Like the SM, SPM conversion is defective in ids1 and rgo1 mutant plants(Fig. 8D). The ids1;rgo1 double mutant phenotype resembles the ids1; ifa1 phenotype in several ways. In the latter, the ear SM is converted into an indeterminate branch meristem which makes SPMs, while the tassel SM is converted into a SPM(Laudencia-Chingcuanco and Hake,2002). Both double mutants have more severe effects in the ear than in the tassel. In both cases, meristem identity has changed and results in a decrease in determinacy. rgo1 and ifa1 do not exhibit a synergistic interaction like ids1 and ifa1 do (data not shown), although ids1 interacts synergistically with bothrgo1 and ifa1. This illustrates that although rgo1and ids1 functions overlap, these genes play unique roles in inflorescence development.

The conversion model of maize inflorescence development proposes that SPMs initiate a SM and then convert into SMs, and SMs intiate a FM and then convert into FMs (Fig. 8A). This model is based on the class II tasselseed phenotypes, in which meristem identity conversions are delayed (Irish,1997a). Determinacy is simply a function of correct identity transitions in the conversion model. A SM that converts into a FM is determinate, while a SM that never changes its identity is indeterminate. An alternative model of inflorescence development, the lateral branching model,does not involve identity conversion (Chuck et al., 1998). Instead, Chuck et al., suggest that all florets,including the terminal floret, are initiated laterally by the SM. A band ofids1 expression between the developing upper and lower floret was interpreted to be the remnant SM, which is not suppressed in ids1mutants. This theory was partially reconciled with the conversion model by Irish who suggested that the remnant of the SM is morphologically indistinguishable but still present on the flank of the FM(Irish, 1998). In the lateral branching model, ids1 functions to suppress indeterminate growth of the SM. One of the predictions of this model is that ids1 mutant SMs should make more FMs, but as SM identity does not change, only FM primordia should be produced by the SM.

Observations from our research favor the conversion model, which can explain mutant behavior without invoking remnant meristems. In the double mutant, SMs produce supernumerary FMs, but also initiate additional glumes and SMs, indicating that SM identity has changed(Fig. 8D). The lateral branching model does not implicate ids1 in meristem identity changes because in this model the SM never becomes a FM. The conversion model explicitly implicates identity change as a mechanism, and thus can explain the double mutant phenotype. The lateral branching model was partially based on the observation that ids1-mum1 spikelets do not terminate in florets,because in this model the florets are lateral branches of the spikelet. Upon introgression into B73, ids1-mum1 plants can make functional terminal florets, as do the ids1-Burr and ids1-VI alleles.ids1-VI ears make kernels and do not exhibit indeterminate spikelets on the ear (Fig. 6). Thus, in certain genetic backgrounds, ids1 mutant spikelets can terminate in fertile florets, suggesting that ids1 is involved in the conversion of SMs into FMs.

Nonallelic noncomplementation can be explained by a genetic pathway that is sensitive to the total dosage of two genes(Stearns and Botstein, 1988;Yook et al., 2001). Based on the nonallelic noncomplementation and synergistic double mutant phenotypes ofrgo1 and ids1, we propose a dosage based model for meristem conversion. In this model, SPM→SM and SM→FM conversions are regulated by both ids1 and rgo1, although the threshold for normal behavior is lower in SPMs than in SMs(Fig. 9). Signals from the vegetative portion of the maize plant are known to affect meristem identity transitions. In the maize indeterminate 1 (id1) mutant the transition from vegetative to reproductive development is extremely delayed or never happens. id1 is expressed in immature leaves, but acts in a non cell-autonomous manner and affects the apical meristem(Colasanti et al., 1998). Meristem culture experiments also support the idea that the SAM integrates signals from the whole plant and does not keep track of its identity internally (Irish and Karlen,1998; Irish and Nelson,1991). If rgo1 and ids1 are in the genetic pathway that responds to these signals coming from the rest of the plant,these mutants could make the meristem less sensitive to the signal needed for meristem conversion (this is equivalent to raising the signaling threshold).

Control of meristem identity transitions in the maize inflorescence is under the combinatorial control of multiple genes. The penetrance of theids1/+; rgo1/+ nonallelic noncomplementation phenotype and the severity of the ids1; rgo1 double mutant phenotype are variable(Fig. 7), suggesting that there are genetic modifiers in the maize genome (and probably environmental effects as well) that affect this phenotype. Three additional lines of experimental evidence provide support for the presence of modifiers. To test the dosage model of nonallelic noncomplementation, we produced plants missing one copy of the long arm of chromosome 1 (which carries ids1) using B centromere translocation lines (Birchler,1996). Chromosome 1L hypoploids exhibited a reversed kernel phenotype. This shows that there are other genes on 1L (possibly tb1,kn1, and/or ts6, which all have effects on meristematic behavior) that in combination with a missing dose of ids1 have an effect on spikelet meristem identity. Similarly, a directed Mutatortransposon mutagenesis experiment aimed at cloning rgo1 produced hundreds of plants with rgo1-like phenotypes. Only one or two disruptions of the rgo1 gene would be expected, suggesting thatrgo1 heterozygotes uncover many branching modifiers. Finally, a large screen of over 50,000 F₁Mutator active ears for dominant or semi-dominant reversed germ phenotypes identified several ears with strong reversed germ phenotypes. These ears transmit fascicled (ears with more than one main spike due to multiple IMs), ramosa (ears with long branches at their base due to indeterminate SPMs), and reversed germ (SM) phenotypes in the F₂ families in a recessive manner. In these families, heterozygotes uncover a reversed germ (i.e. SM) phenotype. However, homozygotes affect both the SM and SPM and IMs. These data suggest there are many genes in the maize genome that, when disrupted, can contribute to the dosage effect that creates reversed kernels.

Several other examples of dosage-sensitive developmental pathways have been described in plants. LEAFY (LFY) is an Arabidopsis gene that regulates the transition to flowering by promoting and maintaining floral meristem identity (Weigel et al.,1992). Heterozygous lfy/+ plants are normal under long day conditions. However, in short day conditions floral reversion results in inflorescences being elaborated from the floral meristem(Okamuro et al., 1996). A second example can be found in the regulation of meristem maintenance. TheCLAVATA1 (CLV1) and CLV3 genes in Arabidopsis are involved in maintaining meristem size by balancing the rates of cell division and primordial initiation. CLV1 encodes a receptor kinase, andCLV3 is its ligand (Trotochaud et al., 2000). CLV1 and CLV3 exhibit a form of dosage sensitivity. The weak clv1/+ phenotype is strongly enhanced in a clv3/+ background (Clark et al.,1995). In addition, stm/+ heterozygotes suppressclv1 semidominance (Clark et al.,1996). Finally, hemizygosity of either liguleless1 orliguleless2, genes necessary for ligule development in maize, affect the null phenotype of the other mutant(Harper and Freeling, 1996). These examples may reflect a general sensitivity to dosage in many aspects of meristem function.

Kellogg (Kellogg, 2000) has proposed a generalized model for inflorescence development. The model states that meristems can produce either more meristems or a set of determinate floral organs on their flanks. In a reiterative manner, any subsequent meristems must make the same developmental decision. This simple model can explain the wide variety of inflorescence forms found in plants. If our results can be generalized, then progress through the various orders of meristem identity will depend on the dosage of ids1 andrgo1-like genes and the concentration of their products. Meristem identity is established by varying a threshold, with different thresholds for different identities. Duplication or loss of these genes would explain variations in the number of iterations of this basic process, and thus could be responsible for the great diversity in inflorescence forms.

We thank Christopher Turbeville for his help with cytology and the staff of the Electron Microscopy Laboratory for their expert assistance. We are grateful to the staff of the Oxford Tract facility, for managing the greenhouse and fields. J. Pfluger, R. Langham, D. Lisch, D. Braun, P. McSteen,G. Chuck and S. Hake provided comments and suggestions. Funding was provided by NIH grant 2R01-GM42610. NJK was supported by a NSF pre-doctoral fellowship.