SEMAPHORE1 functions during the regulation of ancestrally duplicated knox genes and polar auxin transport in maize

The precisely regulated expression of homeobox (hox) genes and gene complexes is a fundamental feature of animal development (reviewed by McGinnis and Krumlauf, 1992). Hox proteins are transcription factors containing a homeodomain, an evolutionarily conserved motif comprising 60 amino acids that functions to control pattern formation and cell specification during development (Laughon and Scott, 1984; McGinnis et al., 1984). First discovered by molecular genetic analyses of homeotic (segment identity) mutants in Drosophila melanogaster, evolutionarily-conserved relatives of the homeobox have since been identified in widespread phylogenetic groups including vertebrates, invertebrates, yeasts and plants (Murtha et al., 1991; Patel et al., 1991; Vollbrecht et al., 1991; Ruberti et al., 1991). Moreover, the residual homology of the metazoan homeodomain to transcriptional regulatory proteins from bacteria and phage lambda suggests that the role of hox genes during development has evolved subsequent to a primordial role in DNA recognition and binding (Laughon and Scott, 1984; Pabo and Sauer, 1984). In lieu of the extended evolutionary history of hox-like genes, the pathways regulating homeobox gene expression may be extremely variable across disparate taxonomic groups.

The maize gene knotted1 is the first homoebox gene identified in plants (Vollbrecht et al., 1991). Dominant, over-expressing alleles of Knotted1 (Smith et al., 1992) and related class 1 KNotted1-like homeobOX (Knox) (Kerstetter et al., 1994) genes in maize (Freeling, 1992; Schneeberger et al., 1995; Muehlbauer et al., 1999; Foster et al., 1999), barley (Müller et al., 1995), rice (Matsuoka et al., 1993), Arabidopsis (Lincoln et al., 1994; Chuck et al., 1996), tobacco (Sinha et al., 1993; Sakamoto et al., 2001) and tomato (Haraven et al., 1996; Chen et al., 1997) condition seemingly disparate, species-specific phenotypes including ectopic meristem formation, delayed cell-fate acquisition, indeterminate growth patterns, increased leaf lobing and super-compound leaf morphology. Null mutations of knox genes may abort the development and/or maintenance of shoot apical meristems (SAMs) (Long et al., 1996; Kerstetter et al., 1997; Vollbrecht et al., 2000), organogenic centers that initiate lateral organs in plant shoots.

The structure of plant meristems is correlated with their function. The central zone (CZ) of the SAM is occupied by large, slowly dividing stem cells, which are surrounded by a peripheral zone (PZ) of smaller, mitotically active cells. Plant lateral organs are formed from founder cells recruited from the peripheral zone (PZ) of the SAM. Anlagen lost from the PZ during organogenesis are replenished from stem cells supplied by the CZ. Thus, the SAM maintains an equilibrium during the production of each plant segment, balancing the production of stem cells in the CZ with the loss of founder cells from the PZ (reviewed by Fletcher and Meyerowitz, 2000). In plants with simple (undissected) leaves, KNOX proteins accumulate in the CZ, but are excluded from lateral organ primordia and also from the organogenic PZ. Patterns of knox gene expression in the SAM support a model whereby KNOX proteins function to promote developmental indeterminacy in plant cells. Consequently, the precise, epigenetic regulation of knox gene expression affects cell-fate acquisition and lateral organ formation during plant development.

The MYB-domain protein ROUGH SHEATH2 (RS2) is the first regulator of knox gene expression found in plants (Schneeberger et al., 1998; Timmermans et al., 1999; Tsiantis et al., 1999b). Recessive mutations in the maize rs2 gene result in the ectopic expression of at least three knox genes (liguless3, knotted1 and rough sheath1) in shoot lateral organs. Correlated with the mis-regulation of these knox genes is an array of mutant shoot phenotypes; many of which mimic those conferred by dominant, neomorphic mutations in knox genes such as rough sheath1 and knotted1 (reviewed by Freeling, 1992). Orthologous mutations are described in myb genes from Arabidopsis (asymmetric leaves1, as1) (Ori et al., 2000; Byrne et al., 2000) and Antirrhinum (phantastica1, phan1) (Waites and Hudson, 1995; Waites et al., 1998). The expression patterns of the rs2/phan/as1 gene family are complimentary to those of knox genes, consistent with their role during negative regulation of knox gene expression. Furthermore, rs2 mutants are defective in polar auxin transport (PAT) (Tsiantis et al., 1999b), which may account for many morphological abnormalities of rs2 mutant shoots.

Another intriguing finding is that although the rs2/as1/phan1 mutations all cause ectopic knox expression, their mutant phenotypes are not fully analogous (Schneeberger et al., 1998; Ori et al., 2000; Waites and Hudson, 1995). Differential defects in proximal-distal (enhanced in rs2), dorsal-ventral (enhanced in phan1), and mediolateral patterning (enhanced in as1) are noted in the three mutants. The variable phenotypes of knox regulatory mutants observed in maize, Antirrhinum and Arabidopsis may be attributed to differences in the specific knox gene targets, differential expression patterns of knox and downstream genes, and/or divergent strategies employed during development of eudicot and monocot leaf morphology (Timmermans et al., 1999; Tsiantis et al., 1999b; Ori et al., 2000).

The accumulated data suggest a model whereby the RS2/PHAN1/AS1 family regulates knox gene expression via the epigenetic control of transcriptional memory in differentiating cells (Ori et al., 2000) (reviewed by Byrne et al., 2001). In support of this model, mutations such as pickel1 and serrate1, both of which affect genes predicted to encode chromatin remodeling factors in Arabidopsis, enhance the phenotypes of knox regulatory mutants but do not themselves condition knox gene mis-expression (Ori et al., 2000). A similar scenario is reported for the handlebars1 mutant (Waites and Hudson, 2001), which greatly enhances the phan1 mutant phenotype. Ultimately, these data suggest that the epigenetic control of knox gene expression during plant development is a complex interaction of multiple genes and gene networks, many of which exhibit phase-specific, organ specific, as well as organ-domain specific effects.

Here we report the function of SEMAPHORE1 (SEM1), the second regulator of knox gene expression identified in maize. The unique and pleiotropic semaphore1 mutant phenotypes include abnormalities in the development of the embryo and endosperm, roots, stem, leaves, vasculature and flowers. These mutant phenotypes are correlated with the mis-expression of the ancestrally duplicated knox genes (Kerstetter et al., 1994; Foster et al., 1999) rough sheath1 and gnarley1, and also with impaired polar auxin transport in semaphore1 mutant shoots. The phenotypic, molecular and genetic data suggest that RS2 and SEM1 function in separate pathways to control knox gene expression in maize shoots. Furthermore, these data support a model in which suppression of knox gene expression promotes basipetal transport of auxin in maize leaves.

The sem1-R mutant (previously dek*Mu1364) was first identified as a recessive, defective kernel mutant phenotype obtained from Robertson’s Mutator stocks (Scanlon et al., 1994). The mutation is introgressed 6 times into the inbred B73, and 5 times into the Q66 inbred line. The genetically unlinked rs2-R mutation (isolated by M. I. Hajidov in 1937) was donated by P. Becraft (Iowa State University).

The sem1*Mu12 allele was obtained via transposon-tagging with Robertson’s Mutator (Robertson, 1978). Homozygous sem1-R plants are male sterile; therefore sem1-R heterozygous males were crossed to plants containing Mutator transposon activity. Among approximately 25,000 F₁ kernels, 18 were selected on the basis of their small endosperm/small embryo phenotype, planted, and grown to maturity. A single plant among the 18, designated as sem1*Mu12, exhibited a small stature and displaced ligule phenotype, and was propagated by out-crossing to inbred W23. The sem1*Mu9 allele was obtained by directed transposon-tagging utilizing plants homozygous for the sem1-R mutation as female. Mutant sem1-R plants were de-tasseled to eliminate self-contamination, and crossed to pollen from Mu-active plants. Among approximately 2,000 F₁ progeny kernels, two were selected on the basis of the small endosperm/small embryo phenotype. The selected seed were sown, and a single plant, designated sem1*Mu9, which displayed the small stature/displaced ligule phenotype was out-crossed to inbred W23.

In order to locate the sem1-R mutation to a particular chromosome arm, a series of plants heterozygous for B/A translocations (Beckett, 1994) (for details, see Scanlon et al., 2000) were crossed as male to plants heterozygous for sem1-R. B/A translocations are useful for genetic mapping (reviewed by Beckett, 1994), and in analyses of gene dosage (reviewed by Birchler, 1994). B/A translocations are reciprocal exchanges between a normal (A) maize chromosome and a supernumerary (B) chromosome. The centromere of the supernumerary B chromosome inherently undergoes non-disjunction at high frequency during the second mitotic division of microsporogenesis. Therefore, plants harboring a B/A translocation typically produce discordant pollen grains, in which the two sperm nuclei are unbalanced with respect to the dosage of genes linked to the B centromere. Specifically, one sperm nucleus contains two doses of genes linked to the B centromere (hyperploid) whereas the other sperm nucleus is deficient (hypoploid). Following double fertilization by a male plant carrying a B/A translocation, discordant aneuploid kernels are generated in which the embryo may be hypoploid and the endosperm hyperploid, and vice versa. In this way, recessive mutations located distal to the breakpoint in the A chromosome will be “uncovered” in hypoploid F₁ progeny.

F₁ progeny of crosses between sem1-R/B73 heterozygotes and male plants hyperploid for TB-9Sd (which carried the linked anthocyanin marker C1) segregated kernels that were discordant for sem1 mutant phenotypes (Fig. 1C). Two discordant kernel phenotypes were noted: (1) kernels with hypoploid, colorless sem1 mutant embryos (genotype sem1-R/–) and purple, non-mutant hyperploid endosperm (genotype sem1-R/sem1-R/Sem1/Sem1); and (2) kernels with purple, non-mutant hyperploid embryos (genotype sem1-R/Sem1/Sem1) and colorless, hypoploid mutant endosperm (genotype sem1-R/sem1-R/–).

Allele tests of sem1-R and both mutant isolates, sem1*Mu12 and sem1*Mu9, were performed by reciprocal crosses between heterozygous plants as described previously (Scanlon et al., 1994). Newly identified sem1 mutants are each expected to be heterozygous for both the sem1-R allele and the newly tagged sem1*Mu allele. Following out-crossing to non-mutant standard lines, the progeny of the newly tagged mutant plants are predicted to segregate 1:1 for heterozygosity for either sem1-R or the newly tagged sem1*Mu allele. RFLP analyses using a genetic marker tightly linked to sem1-R (HN900, no recombinants have been detected among over 250 progeny, unpublished data) were used to distinguish plants heterozygous for the newly tagged mutant sem1 alleles from those harboring the reference allele sem1-R. Positive allelism tests were indicated by the segregation of approximately one in four mutant kernels on ears test-crossed to sem1-R/B73 heterozygous plants.

For the construction of semaphore1-R/rough sheath2-R double mutants, plants heterozygous for the sem1-R mutation were crossed to plants heterozygous for the unlinked rs2-R mutation, and the F₁ plants were self-pollinated. The resulting F₂ progeny were scored for the presence of one in four sem1 mutant kernels. F₂sem1 mutant kernels obtained from self-pollinated ears that also segregated approximately one in four rs2 mutant seedlings were screened for segregation of putative sem/rs2 double mutants. The genotypes of rs2/sem1 double mutants identified in the progeny were verified by DNA gel-blot analyses using a 0.5 kb fragment of the rs2 gene (gift from M. Timmermans), and the sem1-R linked RFLP marker HN900 as probes.

Immunohistolocalization of KNOX proteins in 14-day post-germination maize seedlings and in maize seeds collected 12 days after pollination were performed as described previously (Scanlon et al., 1996) using the KNOX polyclonal antibody (gift from R. Schneeberger). Hypocotyl samples were dissected from sem1-R mutant and non-mutant seedlings (5 samples each), fixed in FAA, paraffin-embedded, sectioned and stained in Safranin-Fast Green as described (Sylvester and Ruzin, 1994).

Leaves were cleared of cytoplasmic components as described by Sylvester and Ruzin (Sylvester and Ruzin, 1994). Harvested leaves were treated in 5% NaOH for 24 hours. Following 2 rinses in distilled H₂O, leaves were incubated in chloral hydrate (250 g/100 ml) overnight. Cleared leaves were stained in 1% saffranin, destained in xylene, mounted in Permount™ (Fisher Scientific) and photographed.

Analyses of polar auxin transport were performed on hypocotyls harvested from dark grown, 5 day-old non-mutant and semaphore1 mutant seedlings as described previously (Tsiantis et al., 1999), with the following modifications. Assays were performed by incubating 2.5 cm segments of hypocotyls oriented base down (acropetal transport) or base up (basipetal transport) in 60 μl of nutrient solution (Gil et al., 2001) containing 12.5 μCi of tritiated indol-acetic acid [³H]IAA for 16 hours. Measurements were performed on 2 mm tip segments taken from 5 hypocotyl segments for each treatment, using a LKB Wallac 1218 Rackbeta scintillation counter.

Scanning electron microscopy (SEM) was performed on epidermal replicas of mature maize leaves and developing leaf primordia as described previously (Sylvester et al., 1990).

Semi-quantitative RT-PCR analyses were performed on cDNA prepared from non-mutant, sem1-R mutant and rs2-R mutant seedlings leaves as described by Bauer et al. (Bauer et al., 1994). Gene-specific primers for rough sheath1, knotted1, liguless3 and ubiquitin were the same as those employed by Schneeberger et al. (Schneeberger et al., 1998). Gene-specific primer sequences for gnarley1 (knox4) provided by J. Yamaguchi and S. Hake (USDA, Albany, CA) were as follows:

knox4-A 5′-tatcgcctcggttcatga-3′

knox4-B 5′-ccgcaccccgccttggtt-3′

First-strand cDNA was prepared from RNA extracted from seedling leaves (dissected away from internode and shoot meristems tissue) or the ligule/auricle region of unexpanded, adult leaves. RT-PCR was performed using gene-specific primers for 25, 30 and 40 cycles (cycle=94°C for 30 seconds; 60°C for 30 seconds; 72°C for 60 seconds) and the products were analyzed by gel-electrophoresis in 1.5% agarose. In order to control for discrepancies in gel mobility or ethidium bromide staining, all PCR products shown in Fig. 4C were loaded onto the same agarose gel. The specificity of the knox gene primers was verified by sequencing cloned PCR products.

The sem1-R mutation was identified in a screen for defective kernel phenotypes derived from maize Mutator-transposon backgrounds (Scanlon et al., 1994). Plants heterozygous for sem1 mutations segregate approximately one to four mutant kernel phenotypes in self-pollinated ears (Fig. 1A). Homozygous sem1-R mutant kernels have both a smaller endosperm and embryo than non-mutant sibling seed (Fig. 1B), and kernel viability is dependent upon genetic background. Following six generations of introgression into the maize inbred B73 approximately 70% of sem1-R mutant kernels are viable. However, five introgressions into the inbred Q66 result in greater than 90% mortality of sem1-R homozygous seed. Included among the lethal sem1-R mutant phenotypes are kernels that fail to germinate, as well as seedling-lethals that develop an abnormal fused coleoptile (described below) but no leaves.

The maize B/A translocations (described in Materials and Methods) (reviewed by Birchler, 1994; Beckett, 1994) were used to locate the sem1-R mutation to the short arm of chromosome 9 (Fig. 1C). Complementation analyses with previously described seed mutations in this map location detected no allelism. Thus, semaphore1 denotes a novel, defective-kernel maize mutation located on the short arm of chromosome 9.

Immunohistochemical analyses were performed on semaphore mutant seedlings using a polyclonal antibody that recognizes class 1 KNotted1-like homeobOX (KNOX) proteins (Schneeberger et al., 1998). KNOX proteins accumulate in the shoot apical meristem (SAM) and stems of non-mutant maize seedlings, but are not detected in leaf primordia or founder cells of the incipient leaf (Fig. 2A,D). In contrast, viable sem1 mutant seedlings exhibit ectopic accumulation of KNOX protein(s) in leaf primordia, predominately in the epidermis (Fig. 2B). Interestingly, down-regulation of KNOX accumulation appears normal in the incipient mutant leaf (Fig. 2E), implying that leaf founder-cell recruitment is not disrupted in viable sem1-R mutants. Furthermore, the aberrant accumulation of KNOX protein(s) is absent from younger mutant primordia (P1-P3), but increases in later stages of leaf development (P4-P5 and older). These data indicate that SEMAPHORE is not essential for repression of KNOX protein accumulation during early stages of leaf development.

Seedling-lethal sem1-R mutants display a novel phenotype (Fig. 2C,F). The first leaf-like lateral organ of non-mutant maize shoots is the coleoptile, a hollow tube that encloses and protects the young leaf primordia during germination. In severe sem1-R mutants, however, the coleoptile develops as a fused shaft that accumulates high levels of KNOX protein throughout. Moreover, seedling-lethal sem1-R mutants produce no lateral organs subsequent to this mutant coleoptile, and no organized SAM is present in the mutant apex (Fig. 2F). This extreme phenotype is more prevalent in the Q66 genetic background, although seedling-lethal phenotypes are seen in <10% of sem1-R mutant seed introgressed into inbred B73. In summary, immunohistolocalization data indicate that SEMAPHORE1 is required for negative regulation of KNOX accumulation in maize seedlings.

The phenotypic variation seen in sem1-R homozygous seedlings is reflected in sem1-R mutant embryos harvested at 12 days after pollination (DAP). Among 10 individual non-mutant sibling embryos examined at 12 DAP all had progressed to stage 2 (Abbe and Stein, 1954), during which two leaf primordia are enclosed within the coleoptile (Fig. 3A,B). In contrast, two distinct classes of embryo phenotype were observed in 12 DAP sem1-R mutant kernels. Moreover, all sem1-R mutant embryos were developmentally retarded. Specifically, seven out of 12 mutant embryos had reached stage 1, during which the coleoptile surrounds a single leaf primordium (Fig. 3C,D). Intriguingly, five of 12 mutant embryos had progressed only to the coleoptilar stage (Fig. 3E-H). Moreover two of these severe mutants displayed aberrant embryo morphology, including an enlarged coleoptile and a broad apex that does not exhibit normal tunica-corpus morphology (Fig. 3G,H). These results suggest that developmentally aberrant embryos such as those shown in Fig. 3G,H may develop into lethal seedlings with abnormal coleoptile and SAM morphology, as shown in Fig. 2C,F.

Immunolocalization studies of maize kernels reveal that KNOX proteins accumulate in sem1 mutant endosperm at 12 DAP (Fig. 3C,E,G), but not in non-mutant endosperm (Fig. 3A) (Smith et al., 1995). Also, the degree of KNOX accumulation in the endosperm is correlated with the severity of the mutant embryo phenotype; extremely retarded mutant embryos (Fig. 3E,G) contain more immunopositive endosperm nuclei than mutant siblings with milder embryo phenotypes (Fig. 3C). However, no abnormal KNOX accumulation is detected in lateral organs of 12 DAP sem1 mutant embryos (Fig. 3). As in non-mutant siblings, sem1 mutants accumulate high levels of KNOX in the shoot apex and embryonic stem, but do not accumulate KNOX proteins in the mutant scutellum, coleoptile or developing leaf primordium. These data are consistent with KNOX accumulation patterns observed in semaphore mutant seedlings; no ectopic KNOX proteins are observed in young, mutant leaf primordia.

The KNOX antibody is not gene specific (Scanlon et al., 1996). Therefore, RT-PCR was used to determine the relative abundance of specific, knox gene transcripts in sem1-R mutant, rs2-R mutant and non-mutant leaves (Fig. 4). Primers specific for the maize ubiquitin gene were used to control for approximately equal concentrations of cDNA in the RT-PCR reactions. RT-PCR reveals that ectopic transcripts of gnarley1 (gn1), and to a lesser degree rough sheath1 (rs1) accumulate in unexpanded leaves extracted from the whorl of young adult mutant plants (Fig. 4A,B). No knotted1 (kn1) RT-PCR products are detected in adult sem mutant leaves (data not shown), whereas no rs1-, gn1-, or kn1-specific RT-PCR products are amplified in non-mutant adult leaves (Fig. 4A,B). Inconsistent results are obtained in our RT-PCR analyses using lg3 primers (Fig. 4B), such that weakly amplified transcripts are occasionally detected in both non-mutant and sem1-R mutant leaves. These observations are in agreement with previous RT-PCR analyses of non-mutant leaves utilizing lg3 primers (G. Muehlbauer, personal communication). Time-course RT-PCR (Bauer et al., 1994) was used to compare knox gene transcript accumulation in cDNA prepared from sem1-R, rs2-R and non-mutant B73 seedling leaves (Fig. 4C). Transcripts from both gn1 and rs1 are amplified from sem1-R seedling leaves following 40 and 30 PCR cycles respectively (Fig. 4C top). Therefore, whereas ectopic gn1 transcripts predominate over rs1 transcripts in sem1-R adult leaves, rs1 is more abundant in sem1-R seedlings. In comparison, abundant transcripts of rs1 are amplified from rs2-R leaves after just 25 cycles, whereas gn1 amplicons are first detected after 30 cycles (Fig. 4C middle). In addition, kn1 transcripts are weakly amplified following 30-40 cycles of RT-PCR in rs2 seedlings, whereas no kn1 is detected in sem1-R seedlings (Fig. 4C). Thus, sem1-R mutant seedlings exhibit a slightly different profile of ectopic knox expression than rs2-R, the first-identified knox-regulatory mutant of maize (Schneeberger et al., 1998). Both mutants mis-regulate rs1 and gn1, whereas ectopic expression of kn1 is detected in rs2-R seedlings only. Moreover, the rs2-R shoot phenotypes are more extreme than in viable sem1-R seedlings, and are correlated with higher levels of ectopic knox expression (Fig. 4C).

After introgression into the inbred B73 genetic background, the sem1-R mutation conditions specific abnormalities in shoot development. Homozygous sem1-R mutant plants have short internodes (brachytic), although they develop the same number of leaves as non-mutant sibling plants (Fig. 5A). Non-mutant maize leaves comprise several domain-specific structures, which display distinct epidermal cell morphologies (Sylvester et al., 1990). The proximal sheath surrounds the stem and is separated from the distal blade by the wedge-shaped auricle and the ligule, an epidermally derived fringe of tissue (Fig. 5B, left). The sem1-R mutant ligule is displaced distally, but only along the midrib (Fig. 5B, right). Multiple ectopic ligules may form distal to the primary ligule, forming a stack of ligules centered over the midrib (Fig. 4B,C). Scanning electron microscopy (Fig. 6) reveals that in contrast to the narrow fringe of ligule that forms in non-mutant leaves, the sem1-R mutant ligule is a broad patch of tissue (Fig. 6E). Furthermore, cell-fate acquisition is altered in-between the stacks of ectopic ligules (Fig. 6F-H), whereas normal midrib cell types are found distal to these disturbed regions (Fig. 6I). This polarized pattern of mutant cell-fate acquisition, in which proximal tissues (i.e. ligule and sheath) are displaced into distal domains (i.e. blade), is a common phenotype in maize mutants that ectopically express knox genes (reviewed by Freeling, 1992). Moreover, the sem1-R ligule phenotype is detectable at very early stages (Fig. 6A,B) [normal ligule development is described by Walsh et al. (Walsh et al., 1998)], suggesting a defect in ligule initiation rather than later-staged, distorted growth along the blade/sheath boundary.

Abnormal tissue proliferation also occurs at the margins of semaphore mutant leaves, near the blade-sheath boundary (Fig. 5D). Ectopic outgrowths of sheath tissue form highly vascularized, symmetrical pairs at the leaf edge. Sheath outgrowths are never observed at only a single leaf margin. Close examination of the sheath outgrowths reveals atypical vascularization patterns. In contrast to the parallel arrangement of major vessels present in non-mutant maize (Sharman, 1942) (and described below), the sem1-R sheath outgrowths are highly reticulated and form large, transverse branch vessels (Fig. 7A). Although the sem1-R sheath outgrowths are hyper-vascularized, overall transverse vein development in the mutant sheath is markedly deficient. In contrast to the thick transverse vessels found in sem1 mutant sheath outgrowths, transverse veins in non-mutant sheaths are thinner than the parallel, lateral veins (Fig. 7B-D). Non-mutant maize leaves form narrow, yet proliferative transverse connections between the larger lateral vessels in the sheath (Fig. 7B-D). Normally, the elaboration of minor, transverse vessels begins almost immediately proximal to the ligule (Fig. 7B); maize leaf blades develop far fewer transverse connections (Sharman, 1942) (and data not shown). However, transverse vessels in the sem1-R mutant leaf sheath are smaller and less abundant than in non-mutant leaves, especially near the sheath margins (Fig. 7E-G). No gross abnormalities are observed in the leaf-blade vasculature of semaphore1 mutants (data not shown).

In addition, viable sem1-R mutant plants are male-sterile. Fertile ears do develop however, such that homozygous mutant plants may be propagated as female. More variable mutant phenotypes are seen in non-introgressed genetic stocks. Additional sem1-R mutant phenotypes may include the formation of ectopic auricle or ligule tissue on leaf margins (Fig. 5E), and extremely narrow leaves (data not shown). Genetic background strongly influences the sem1-R mutant phenotype, which suggests that many unidentified genetic factors contribute to SEMAPHORE1 function, KNOX accumulation, and knox gene regulation.

A prominent phenotype of sem1-R mutant plants is the formation of yellow stripes on vegetative leaves (Fig. 8A). This pattern of leaf-striping is a diagnostic feature of plants deficient in the uptake of soil magnesium (Olsen and Sander, 1988). However, the leaf-striping phenotype of sem mutant leaves is only partially alleviated by magnesium supplementation. Therefore, we suspected that sem1-R mutant plants might have root abnormalities that result in magnesium-deficient plants. When compared to non-mutant siblings at an equivalent stage of shoot development (determined by the total number of leaves produced), sem1-R mutant seedlings had fewer and severely stunted primary, seminal, and adventitious roots (Fig. 8B,C). Furthermore, there is a striking reduction in the development of lateral roots. As such, sem1 represents the first maize knox-expression mutant shown to disrupt both shoot and root development. Transverse sections of the embryonic stem (Fig. 8D-E) of sem1-R and non-mutant seedlings reveal no gross abnormalities in vascular development of the mutant hypocotyl. However, the hypocotyls and vessel elements in the smaller, sem1-R mutant seedlings are thinner in diameter than those of larger, non-mutant siblings.

The phytohormone auxin undergoes basipetal transport in plant shoots, from the tip to the base of the plant (reviewed by Berleth and Sachs, 2001). The major bio-active auxin in plants is indole-acetic acid (IAA). IAA is known to affect an array of plant developmental processes, including embryonic pattern formation, lateral organ initiation, suppression of lateral branching, vascular development and lateral root development. Regulation of IAA function in plant shoots is proposed to depend upon an equilibrium maintained between auxin biosynthesis and polar auxin transport (PAT). Measures of PAT in sem1 and non-mutant shoots reveal a significant reduction in basipetal transport of IAA in mutant hypocotyls (Fig. 9). The ratio of basipetal:acropetal transport of ³H-labeled IAA in non-mutant sibling shoots is approximately 9.7:1, whereas this ratio is reduced to 2.7:1 in sem1 mutant shoots (Fig. 9). Moreover, the vascular development in the sem1-R hypocotyl is not overtly different than non-mutant siblings (Fig. 8D,E). Therefore, the reduction in PAT in sem1-R mutant shoots cannot be attributed solely to aberrant vasculature in sem1-R hypocotyls.

Analyses of genetically discordant kernels generated by the B/A translocation TB-9Sd (described in Materials and Methods) reveal that the sem1 locus maps to the short arm of chromosome 9, distal to the breakpoint of TB-9Sd (Fig. 1C). No class 1 knox genes of maize have been mapped to this genetic location (Kerstetter et al., 1994). Furthermore, homozygous (sem1-R/sem1-R) mutant plants have comparable phenotypes to plants derived from all 23 semaphore1-hypoploid embryos (genotype sem1-R/–) tested in this study. Therefore, plants harboring a single dose (and no non-mutant copies) of the semaphore1-R mutation are phenotypically indistinguishable from plants harboring two doses of sem1-R. These dosage analyses strongly suggest that the sem1-R allele is a null, recessive mutation.

A directed Mutator transposon-tagging strategy was employed (described in Materials and Methods) in order to generate two independently isolated mutants that exhibit sem1-like phenotypes. Subsequent testcrosses verified that two new sem1-allelic mutations, designated sem1-Mu9 and sem1-Mu12, have been identified. Following a single generation of out-crossing, the genetically heterogeneous sem1-Mu9 mutants are predominantly lethal kernel phenotypes; introgression into B73 and Q66 is in progress. At an equally early stage of introgression, viable sem1-Mu12 mutant plants are obtained in low frequency. The mutant phenotype of viable sem1-Mu12 and sem1-Mu9 plants resembles that of the reference mutant, sem1-R.

SEMAPHORE1 represents only the second knox gene regulatory function described in maize. Accordingly, sem1-rs2 double mutant plants were constructed (described in Materials and Methods) in order to test whether SEMAPHORE1 and ROUGH SHEATH2 function in a shared developmental pathway(s). RFLP analyses confirmed the double-mutation genotypes. The leaf phenotypes of rs2 mutants are more extreme than viable sem1-R mutants (this work) (Schneeberger et al., 1998). Moreover, disruptions of the blade/sheath boundary conditioned by the rs2-R mutation are so severe as to mask the displaced ligule phenotype of sem1-R mutant plants. Therefore, analyses of epistasis are difficult in sem1-rs2 double mutants. However, the appearance of numerous sem1-specific phenotypes (embryo and endosperm defects, extreme brachytic stature, ectopic sheath extensions, zebra-striped leaves and male sterility) in combination with rs2-specific leaf phenotypes indicate that the double mutant phenotype is additive. Furthermore, sem1 and rs2 mutants ectopically express slightly different profiles of knox genes. These results suggest that SEMAPHORE1 and ROUGH SHEATH2 may regulate maize knox gene expression in separable developmental genetic pathways.

semaphore1 mutants ectopically express at least two class 1 knox genes, the homologues rough sheath1 and gnarley1 (knox4) (Kertstetter et al., 1994; Schneeberger et al., 1995; Foster et al., 1999). The coding regions of RS1 and GN1 share greater than 90% similarity, and these loci map to duplicated regions of the maize genome. Therefore, SEMAPHORE1 may function to down-regulate the expression of a specific, subset of knox gene homologues. Correspondingly, many of the shoot phenotypes observed in semaphore1 mutant plants are similar to those previously described for dominant mutations in maize knox genes (reviewed by Freeling et al., 1992), as well as the recessive, knox-regulatory mutant rough sheath2 (Schneeberger et al., 1998). In addition, although the leaf phenotypes of sem1 mutants are restricted to specific domains of the midrib and margins (Fig. 5), RT-PCR assays and KNOX immunohistolocalization assays reveal that ectopic knox gene expression transcends these domains and includes leaf regions that exhibit no mutant phenotype (Fig. 2 and data not shown). These results concur with those reported for dominant knox mutations such as Rs1 and Lg3, and further suggest that the phenotypic consequences of knox gene expression in maize leaf cells may be influenced by their competency to respond to ectopic signals (Schneeberger et al., 1995; Muehlbauer et al., 1997).

Moreover, KNOX down-regulation is normal in the founder cells of sem1 mutant leaves; ectopic KNOX accumulation is not observed until later primordial stages of leaf development (Fig. 2B,E). These data imply that the initial down-regulation of KNOX accumulation in maize founder cells and young leaf primordia is controlled by separate gene function(s), whereas SEMAPHORE1 is required to maintain knox transcriptional repression during later stages of lateral organ development. An intriguing aspect of the semaphore1 mutation is the phenotypic variation of semaphore homozygotes. Even in introgressed genetic backgrounds, homozygous sem1 mutant phenotypes range from seedling/embryo-lethals to viable, more-moderately affected plants. In less-introgressed genetic backgrounds, phenotypic variability is summarily increased (Fig. 5E). Furthermore, dosage analyses reveal that sem1-R is a null allele (Fig. 1C), indicating that this phenotypic variation is not attributable to a ‘leaky’ gene product. Phenotypic variability is also noted in knox-regulatory null mutants of the phan1/as1/rs2 family (Waites and Hudson, 1995; Ori et al., 2000; Schneeberger et al., 1998), as well as the maize dominant, neomorphs Kn1, Rs1, Lg3 and Gn1 (Freeling, 1992). These data indicate that complex epigenetic pathways interact during the regulation of – and responses to – ectopic knox gene expression in plant shoots (reviewed by Byrne et al., 2001).

Unlike previously described knox mutants in maize, the sem1 phenotype is not restricted to shoot lateral organs. Thus, sem1 presents the first known correlation between ectopic knox gene expression and abnormal endosperm, embryo and root development. Furthermore, the ectopic accumulation of KNOX proteins in sem1 mutant endosperm represents the first observation of ectopic knox expression in this tissue (Fig. 3C,E,G), and further suggests that SEMAPHORE1 function is required to maintain knox transcriptional repression. Interestingly, sem1 mutant embryo development is disrupted prior to the abnormal accumulation of KNOX proteins in shoot lateral organs. These results suggest at least two possible causes of defective patterning in sem1 embryos. Perhaps embryo development in sem1 mutants is hampered by interaction with an ectopic-knox-expressing endosperm? Alternatively, SEMAPHORE1 may have an additional function(s), which is required for embryonic pattern formation.

In addition to the classical ‘knox’ mutant shoot phenotypes described above, sem mutants also exhibit distinct phenotypes that have not been described previously for maize knox expression mutants. These include defects in endosperm development, embryonic patterning, lateral root initiation, vascular development and pollen viability. Mutations at semaphore1 also disrupt PAT (Fig. 9), as do maize rough sheath2 mutations (Tsiantis et al., 1999b). Consequently, it is possible that defective PAT may be responsible for many of the developmental defects conditioned by sem1 mutations. For example, defective basipetal (tip to base) transport of auxin is known to affect internode elongation, lateral root initiation, vascular development and embryonic patterning (reviewed by Berleth and Sachs, 2001). In addition, lethal sem1 seedlings form a fused coleoptile and no SAM (Fig. 2C,F). Likewise, aberrant PAT is associated with abnormal meristem and coleoptile development in wheat embryos (Fischer and Neuhaus, 1996), as well as organ fusion in the cotyledons of Arabidopsis and Brassica (Okada et al., 1991; Bennett et al., 1995; Liu et al., 1993; Hadfi et al., 1998).

Furthermore, vascularization of sem1 mutant leaves is markedly deficient in sheath tissue. In contrast no variations are noted in the mutant leaf blade, whereas hypervascularization is observed in sheath extensions located at the marginal, blade-sheath boundary (Fig. 7). Taken together, these data suggest a model in which defective PAT in the sheath may result in localized pooling of auxin concentrations at the blade-sheath boundary. Moreover, local maxima of auxin concentrations may accumulate along the vasculature of the midrib and margin domains, putative routes of auxin transport in shoots (Berleth and Sachs, 2000). This model may explain the generation of an auxin deficit in the basal regions of the phytomer, perhaps resulting in hypovascularization of the sheath, short stems and reduced lateral root formation. Likewise, pooling of auxin over the midrib and margins of the blade/sheath boundary may cause proliferative growth over the midrib and hypervascularized extensions at the margins, all of which are nominal phenotypes of sem1 mutants (Fig. 5).

Numerous reports suggest a link between ectopic knox gene expression and defective regulation of hormone concentrations in plant shoots. In particular, knox over-expression results in elevated levels of cytokinin (Frugis et al., 1999; Ori et al., 1999), whereas the KNOX protein NTH15 has been shown to directly suppress the accumulation of gibberellin in the tobacco SAM (Sakamoto et al., 2001). Moreover, recessive mutations in the knox regulator rs2 are correlated with defective auxin transport in maize shoots (Tsiantis et al., 1999a). Tsiantis et al. also showed that chemically induced disruptions of PAT induce rs2-like ‘knox’ phenotypes in maize seedlings. Intriguingly, no ectopic knox gene expression is detected in maize seedlings treated with PAT inhibitors (Tsiantis et al., 1999a); these data suggest that defective PAT is a downstream effect of ectopic knox expression, rather than vice-versa. The sem1 mutant phenotypes provide further evidence that the regulation of knox gene expression impacts plant hormone levels.

We wish to thank P. Becraft for rs2-R mutant seed stocks; M. Timmermans for the rs2 gene-specific probe; D.S. Robertson and P. Stinard for the initial donation of B-A translocation mapping stocks and J. Yamaguchi and S. Hake for knox4-specific primer sequences. We thank A. Myers, D. Robertson and M. Freeling for support during early stages of this work. Thanks to the S. Wessler and K. Dawe laboratories for robust discussions of these data. This work is funded by USDA grant 99-35304-8096 to M. S.