Genetic approaches to inflorescence and leaf development in maize

Although it is clear that development is guided to a large extent by information encoded in genes, describing how this is accomplished has been challenging. A detailed description of the process through developmental studies is an important first step, but this approach by itself cannot resolve the activity of individual genes. Molecular studies, by contrast, succeed well at describing gene products, but fail to demonstrate specific functions for them in the larger process.

Genetics affords a solution to this problem by laying a phenomenological framework that is relevant in both developmental and molecular contexts. We describe here several examples that illustrate our use of genetics to understand morphogenesis in maize. We begin with floral development to show how genetics can help define relationships between individual components of a complex process. The focus then shifts to leaf development with discussion of how dosage and mosaic analyses have clarified the nature of the Knotted (Kn1) gene and provided a framework for molecular studies. Finally we consider strategies for integrating molecular data into genetically grounded models of development.

Several aspects of the maize inflorescence make it a convenient system to study morphogenesis. Although the development of the inflorescence is a complex process, it can be readily divided into a sequence of morphologically distinct stages. Because the inflorescence is a reiterated structure, with the oldest, most differentiated florets at its base, departures from the normal sequence of development are easily discerned.

Finally, mutants exist that alter specific aspects of inflorescence development (Sheridan, 1989; Coe et al. 1988).

Normal development can be divided into two broad phases, as suggested by the classic studies of Bonnett (1940,1948) and the later more refined SEM analysis of Cheng et al. (1983). The first begins with the transformation of a vegetative meristem into a floral axis (Fig. 1A). After the last leaf is initiated, the meristem elongates with spikelet pair primordia forming in an acropetal sequence. These primordia divide to form two individual spikelet initials, each of which initiates a pair of flanking glumes. The spikelet primordia then divide to give upper and lower floret primordia. Each floret axis then produces, in an acropetal sequence, the initials for the glume-like lemma and palea, the stamens and lodicules, and finally, the centrally located pistil. At this point, the immature ear and tassel are bisexual and essentially isomorphic structures that are distinguished primarily by their respective axillary and terminal positions.

During the second phase of development, the tassel (Fig. IB) and ear (Fig. 1C) become morphologically distinct. The staminate florets of the tassel and the pistillate florets of the ear differentiate by the respective abortion of pistil and stamen primordia. The lower florets of ear spikelets typically disintegrate, whereas they remain in tassel spikelets. In most lines of maize, tassels develop spikelet-bearing lateral branches.

A phenotypic analysis of inflorescence mutants hints at some of the basic logic of floral development. The vegetative growth of most of these mutants is relatively normal, suggesting that the corresponding genes are required only during flowering. Inflorescence-specific mutants can be grouped in a hierarchy. Mutants such as silkless (sk) (Jones, 1925) or Vestigial glume (Vg) (Sprague, 1939), each of which alters only one type of inflorescence structure, suggest that gene products pertain exclusively to the development of a single organ. Other mutants show transformations between organ series, suggesting a class of genes that determine organ identity. For example, in the mutant silky (si), pistil-like structures develop in place of anthers in the ear (Fraser, 1933; B. Veit, unpublished observations). A third class of mutants defines genes that coordinate the development of more than one organ series. For example, mutants such as ts2 (tasselseed) convert the tassel into a kernel-bearing structure by reversing the normal pattern of organ abortion in the tassel. (Emerson, 1920; Nickerson and Dale, 1955; Irish and Nelson, 1989; B. Veit, unpublished observations).

Simple comparisons between mutant and normal patterns of development can suggest the time when a gene is required. Postlethwaite and Nelson (1964) discriminated between several early acting inflorescence mutants, based on the stage at which their phenotypes were first visible. They suggested that such mutants might define critical ‘switch points’ in development, with normal gene function required to progress to subsequent steps in a developmental pathway. They proposed, for example, that the ral (ramosal’, Beadle, 1932) mutation interferes with an early decision in which spikelet pair primordia are determined, causing their replacement with less determinate branches.

We have applied this comparative approach to other mutants to make provisional assessments of when the corresponding genes are normally required. This method cannot determine the precise time of gene action since there will always be some undefined lag before differences in gene expression are manifest on a morphological level. Nevertheless, it does define a late limit for the time of gene action. One example (Fig. 2) compares the sequence of development of a normal ear with that conditioned by the mutant bd (branched silkless;Kempton, 1934). Development proceeds normally up to the point when floret primordia are formed. Although each spikelet gives rise to what appear to be the upper and lower floret primordia, an indeterminate number of abnormal divisions ensue to produce a highly branched structure. Thus, in contrast to ra1, which interferes with determination of the spikelet, bd appears to block determination of the floret.

Tests of epistasis have proven useful in defining functional relationships between mutationally defined genes. While the validity of this approach has been most clearly demonstrated in simple microbial systems (reviewed by Botstein and Maurer, 1982), the predictive value of epistasis studies can be seen in examples of Drosophila and Caenorhabditis development (Nusslein-Vollhard and Wieshaus, 1980; Baker and Belote, 1983; Hodgkin, 1990). The utility of this method in the analysis of plant developmental mutants has been demonstrated in a number of studies (Marx, 1987; Bowman et al. 1989). We describe here two inflorescence mutant combinations that illustrate the logic of the approach.

The first simple case examines the combination of ts2 and Miniplant (Mpl). ts2 conditions the development of a completely pistillate tassel. Mpl (Harberd and Freeling, 1987), a dominant dwarf mutant, has relatively little effect on the tassel, but instead blocks abortion of stamen primordia in the developing ear to give perfect florets (anther-ear). The double mutant is an anther-eared dwarf with perfect florets in the tassel (Fig. 3). Neither mutant perturbs the expression of the other (i.e. the double mutant phenotype is a superimposition of those phenotypes associated with the single mutants), suggesting the genes affect pathways that proceed independently of each other.

The second case considers the double mutant combination of ts2 and sk (silkless), two genes that appear to interact with each other. If these mutants result from loss of function, as suggested by their recessive character, the wild-type Silkless (Sk) gene could be thought of promoting formation of silks in the ear, whereas the wild-type Tasselseed2 gene ( Ts2) would act to suppress the formation of silks in the tassel.

Two alternative models would relate the activities of Ts2 and Sk (Fig. 4). The first supposes that Sk has the potential to promote silk development in both the ear and tassel, but is suppressed in the latter by the activity of the Ts2 gene. The model supposes that Ts2 is normally not expressed in the ear, leaving the silkpromoting Sk gene unsuppressed. An alternative model supposes that Ts2 has the potential to ‘suppress silk development in both the ear and tassel, but is suppressed in the ear by Sk. While both models account equally well for the single mutant phenotypes, only the second correctly predicts the observed tasselseed phenotype of the double mutant (Jones, 1932, 1934).

The long term aim of this approach is to develop internally consistent models for gene interactions by examining many double mutant combinations. Studies that are in progress have so far revealed relatively few cases of epistasis. With a few exceptions, such as ts2 and sk. the majority of double mutants show additivity of mutant phenotypes, especially in those cases where mutants affect the development of different organs. Nonetheless, these data should provide a useful foundation for molecular studies of inflorescence development.

In contrast to the preceding analysis, which aims to describe temporal and functional relationships between genes, we have also approached development by focusing on a single gene, Knotted (Kn1), that affects leaf development. The normal pattern of leaf development has been described in a number of studies (Sharman, 1942; Dale, 1988; Langdale et al. 1989; Nelson and Langdale, 1989). In 1922, Kn1-O, a mutant with hollow, finger-like outpocketings of tissue on its leaves, was discovered in a row of Walden Dent corn. The Kn1 locus is now defined by several additional dominant mutations which are considered alleles to the extent that they share similar phenotypes and a common map position, near Adh1on the long arm of chromosome 1 (Freeling and Hake, 1985).

Although knots conditioned by Kn1 can lend a grotesque appearance to affected plants (Fig. 5), it would be wrong to dismiss them as ‘teratologies’. On the contrary, several aspects of the Knl phenotype display a positional specificity that suggest the Kn1 gene participates in the process of normal leaf development. For example, knots form on the blade portion of the leaf, and are found much less frequently on the more proximal sheath (Fig. 6). Within the blade, the knots are centered over the lateral veins. These veins, even when lacking knots, show a relatively consistent thickening in which parenchymatous cells replace the bundle sheath and associated upper sclerenchyma. In contrast, the smaller intermediate veins appear normal. Further specificity is seen in the characteristic differences that characterize different Kn1 alleles (Freeling and Hake, 1985). For example, knots conditioned by Knl-0 (O = original) occur most frequently on leaves initiated late in development, while those conditioned by Knl-2F11 tend to occur on embryonic leaves. The two mutants also differ with respect to displacement of the ligule, a flap of tissue that normally forms at the junction of the blade and sheath (Fig. 6). While displacement is relatively common in Knl-O, with ectopic ligule running parallel to lateral veins, it rarely is seen in Knl-2F11.

Dosage analysis has been used to address the dominant expression pattern that characterizes all Knl mutants (Freeling and Hake, 1985). Knots do not appear to result from reduced expression of a normal gene product since plants which are hypoploid for the long arm of chromosome 1 (which would contain the hypothetical wild type Knl gene) are normal. Conversely, increasing the dose of a smaller 18 map unit segment of chromosome 1 from the normal 2 copies to 4, fails to produce knots, suggesting that Knl conditions the phenotype through expression of a novel gene product or a substantial overexpression of a normal product.

Further insight into the nature of the Kn1 gene product has been gained through mosaic analysis. Clonal sectors of tissue which lacked Kn1 were generated by X-irradiating young plants which were heterozygous for Kn1. Sectors lacking Kn1 are easily recognized by the absence of chlorophyll that results when the closely linked gene Iw (lemon white) on the opposite homologue is uncovered. Pure white sectors were unknotted, suggesting that Kn1 activity does not readily diffuse in a lateral direction (Hake and Freeling, 1986). More refined analyses were performed to ask in which cell layers Kn1 must be expressed for knots to form. Examination of numerous mericlinal and sectorial chimeras in which one or more layers lacked the Kn1 gene indicated that Kn1 need only be present in the internal layers corresponding to the bundle sheath to perturb the development of all cell layers (Sinha and Hake, 1990). Thus, knotting provides an example of induction in plants, in which the developmental fate of one cell layer is determined by a Kn1-dependent signal that originates in an adjacent layer.

The cloning of Knl-2F11 has provided opportunities to test and refine models based on genetic analysis. One such approach seeks to define a structural basis for phenotypic differences that characterize different Knl alleles. Knl-2F11 was cloned by virtue of its association with the previously cloned transposable element, Ds2 (Hake et al. 1989). Moreover, analysis of revenants suggests Ds2 is responsible for the mutant phenotype, since knotting is absent in lines where the Ds2 has transposed away.

Comparison of Knl-2F11 with several Kn1 alleles (designated Kn1-mum) that arose in lines carrying Robertson’s Mutator (Robertson, 1978) reveals some similarities. Like Knl-2F11, the Kn1-mum alleles are associated with transposon insertions (Walko and Hake, manuscript in preparation). The inserts, members of the one Mutator class of transposons, lie near where the Ds2 element is inserted in the Knl-2F11 (Fig. 7A). However, in contrast to the mild knots conditioned by Knl-2F11, Kn1-mum knots are more severe and are found on leaves initiated later in development. Kn1-O, an allele which like the Kn1-mums conditions knots late in development, lacks any sizeable insertion element. Instead, the allele is associated with a tandem duplication of a 17 kb region (Fig. 7B). Knotting is absent in derivatives in which the duplication has been lost or disrupted, while gain of a third repeat copy exacerbates knotting (Veit et al. 1990) .

The sequence alterations associated with various Knl alleles, involving a relatively small region (<2kb), prompted an analysis of transcription (Vollbrecht et al. 1991) . Probes derived from sequences lying between the novel duplication junction of Knl-0 and the Ds2 insertion site of Knl-2F11 detect a single 1.7 kb polyA+ transcript in both Knl-0 and normal plants. Near full length cDNAs corresponding to the transcript were cloned from normal or Knl-0 plants and found to be essentially identical (Fig. 7B). The gene spans 8 kb and includes 4 introns. Primer extension and SI analysis locate the transcribed portion of the gene entirely within the Knl-O repeat unit. The unusually large third intron contains the insertion sites of transposons associated with the Knl-2F11 and Knlmuin alleles (Fig. 7A). A homeodomain sequence motif is found in the fifth exon, suggesting that the Knl gene product is a transcription factor.

Thus far, we have detected no qualitative differences between transcripts expressed in normal vs. Kn1-0 or Knl-2F11 plants. Quantatitive differences are observed only for Kn1-0 in which expression is proportional to gene copy number. While this difference might suggest that the knotted phenotype results from a simple increase in levels of a normal transcript, two lines of evidence argue against this model. First, as shown by previously discussed dosage analysis, doubling the normal dose of Kn1+ with segmental translocations does not produce knots: however, we have not yet discounted the possibility that dosage compensation is effected at the level of transcription by these larger duplications. A transposon mutagenesis study (Veit et al. 1990) casts further doubt on the simple overexpression model. Nine normal plants were found in a screen of 10000 Knl-0 heterozygotes. Of the normals, five had lost one of the tandem duplication copies, while the remaining four had sustained insertions within the tandem duplication. The distribution of insertions appears non-random (Fig. 7C). Although the Knl gene is represented twice in the duplication, all four insertions fall within a 2 kb segment that is immediately 5’ to one of the gene copies. Thus it appears that the two gene copies are not equivalent.

The clustered distribution of insertions that knockout Knl-O suggests that the novel junction it contains, lying immediately 5′ to one of the gene copies, might condition knotting through aberrant expression of a normal transcript. This model is supported by in situ localization studies that show a novel distribution of transcripts in Knl plants. Signal that is normally concentrated in the schlerenchyma around lateral veins is more widely distributed in Kn1 plants, with lower levels of hybridization seen in sclerenchymatous tissue and in the bundle sheath of intermediate veins (Sinha, 1990). By analogy with the enhancer element located within the intron of the immunoglobulin heavy chain gene (Banerji et al. 1983), it is conceivable that similar expression patterns are conditioned by the transposons of Knl-2Fll and Knl-tnum alleles. Alternatively, these transposons might condition knotting by serving as initiation sites for truncated, misexpressed transcripts.

Although the cloning of Knl represents a significant step towards understanding leaf development, a mechanistic explanation for Knl action awaits characterization of other components. Genetics provide some clues to their nature. Many components are likely to correspond to mutants that, like Knl, interfere with normal leaf development. Their relationship to Kn1 is being addressed by approaches similar to those that have been applied to inflorescence development. Evidence for other components can be seen in the form of background effects that mute or enhance the expression of Kn1. Unfortunately, attempts to resolve individual modifiers of Knl have generally been unsuccessful, suggesting a polygenic basis for most of the background differences. One exception, sok (suppressor of Knotted), was resolved genetically as a recessive suppressor of Knl-O (B. Greene, unpublished observations), sok by itself does not appear to condition abnormal leaf development. It is not yet known whether sok acts in an allele-specific manner.

Knl is also suppressed in plants which are monosomie for IL (Freeling and Hake, 1985). Interestingly, suppression is not observed in plants in which Knl is heterozygous with a small deficiency (<2 map units), implying that Knl expression may be modified by specific dosage-sensitive factors on IL. Alternatively, the altered growth characteristics generally associated with hypoploidy might in some way suppress knotting. Tests are in progress to distinguish these models.

Although we are attempting to clone some of these other genes through transposon tagging, the cloning of Knl enables several alternative strategies. One strategy is suggested by the observation that many genes involved in Drosophila larval development share a conserved homeodomain motif (Hoey and Levine, 1988). These and other studies (reviewed by Biggin and Tjian, 1989) suggest a prominent role for structurally related transcription factors in regulating animal development. That plant development may be regulated in a similar manner has been suggested by the presence of a DNA-binding motif in Knl, as well as in two genes that affect floral development, agamous from Arabidopsis (Yanofsky et al. 1990) and deficiens from Antirrhinum (Sommer et al. 1990). Hybridization experiments suggest the existence of a number of such homeodomain genes in maize and other species which we are now characterizing in more detail (B. Lowe, E. Vollbrecht and R. Walko, unpublished observations).

Given that Knl appears to be a transcription factor, we may ask the obvious question of what it binds to. Recent technical innovations have made cloning of such target sequences possible (Sompayrac and Danna, 1990). Similarly, Knl itself is likely to be a target of transcriptional regulation. We are using transient expresssion systems to define cis elements involved in such regulation with the hope they will lead us to trans acting factors (Vinson et al. 1988).

These approaches will almost certainly yield clones, but they say little about a gene beyond it having certain molecular properties that led to its isolation. By contrast, the functions of genes isolated by transposon tagging are often suggested by the altered phenotypes that result from insertions of the tagging elements. In some cases, it may be possible to show that genes isolated by molecular methods correspond to mutationally defined genes whose mutant phenotypes would then be relevant. In maize, a preliminary assessment can be obtained very quickly by mapping with recombinant inbred lines (Burr et al. 1988).

While other types of correlative evidence may also be instructive, the most definitive tests of gene function will require transgenic methods, whereby the consequences and causes of modified expression patterns can be studied directly. Clones that correspond to loss of function mutations can be approached directly through complementation studies. For other clones, dominant phenotypes resulting from artificially induced expression might suggest function. In the case of Knl, itself a dominant, our goal is to resolve those elements that are required to induce the knots by which we came to know the gene.

We thank Dawn Banasiak for assistance with figures and Ron Wells for help with the manuscript preparation. This work was supported by a National Science Foundation grant DMB-88-19325 to S.H. and the US Department of Agriculture CRIS no. 5335 2230 00100D.