INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Formally, cell shape can be considered to be established in three steps (Hülskamp et al., 1998). In a first step, spatial information, e.g. cell polarity, is established by intracellular mechanisms or provided by outer cues. In a second step, this information is used to reorganize the cell, e.g. change the cytoskeletal arrangement. Finally, actual growth takes place, which includes the incorporation of membrane and cell wall material at defined areas of the cell periphery.

Although some of the biochemistry of the last two steps of the cytoskeletal function and cell wall synthesis is known, the mechanisms underlying the spatial control of cell morphogenesis are largely elusive. Single-cell model systems such as pollen tubes, root hairs, and leaf hairs (trichomes) that are accessible to genetic approaches provide the means to study spatial control mechanisms (Aeschbacher et al., 1994; Marks, 1997; Oppenheimer, 1998; Hülskamp et al., 1999; Kost et al., 1999; Wilhelmi and Preuss, 1999). Among these model cell types, trichomes in Arabidopsis are particularly well suited because they consistently develop a complex three-dimensional form, thus providing excellent criteria to isolate mutants affecting discrete aspects of morphogenesis (Oppenheimer, 1998; Hülskamp et al., 1999).

Leaf trichomes in Arabidopsis are large single cells that originate from the epidermis and are up to 500 µm tall. After trichome fate commitment the cells stop dividing but continue DNA synthesis (endoreduplication; Hülskamp et al., 1994). The incipient trichome extends out of the leaf surface and undergoes two successive branching events (Hülskamp et al., 1994). The orientation of the first branching is co-aligned with the proximal-distal leaf axis. The primary branch, which points toward the leaf tip (main stem), undergoes a second branching in a plane perpendicular to the primary branching plane (Folkers et al., 1997). Subsequently, the trichome extensively elongates concomitant with an increase in vacuolization. The mature trichome has, on average, a DNA content of 32 C, suggesting that trichomes proceed through four endoreduplication cycles (Hülskamp et al., 1994).

Trichome branching requires the coordinated action of at least 18 genes (Marks, 1997; Oppenheimer, 1998; Hülskamp et al., 1999). One group of genes appears to affect primarily the number of endoreduplication cycles and probably, as a consequence, also branch number (Hülskamp et al., 1994; Perazza et al., 1999; Kirik et al., 2001). A second group of branching mutants affects branch number without affecting endoreduplication (Folkers et al., 1997; Luo and Oppenheimer, 1999; Qiu et al., 2002). The genetic analysis of branching mutants suggests that branch formation is controlled by several redundant pathways (Luo and Oppenheimer, 1999).

To date, five branching genes have been cloned and all appear to be involved in the regulation of the microtubule cytoskeleton at different levels. The ZWI (ZWICHEL) gene encodes a kinesin motor protein with a calmodulin-binding domain, indicating that microtubule-based transport is important for branch formation (Oppenheimer et al., 1997). That the spatial organization of microtubules is important for trichome branching is suggested by the finding that in an (angustifolia) mutants, reduced trichome branching is correlated with the failure to establish a higher microtubule density at the tip of the developing trichome (Folkers et al., 2002). The underlying biochemical mechanism, however, remains unclear because AN encodes a novel protein with sequence similarity to C-terminal binding protein/BrefeldinA ribosylated substrates that are known to be involved in transcriptional regulation or in vesicle budding but not in microtubule function (Folkers et al., 2002; Kim et al., 2002). The FRA2 (FRAGILE FIBER2)/ERH3 (ECTOPIC ROOT HAIR3) gene appears to be involved in the regulation microtubule assembly and disassembly. In fra2/erh3 mutants trichomes are underbranched; also, other cell types show morphogenesis defects (Burk et al., 2001; Webb et al., 2002). FRA2/ERH3 encodes for a katanin-p60 protein, suggesting that it functions as a microtubule-severing protein (Burk et al., 2001; Webb et al., 2002). In fs (fass)/ton2 (toneau2) mutants, shape changes of various cell types have been correlated with severe distortions of the microtubule cytoskeleton (Traas et al., 1995; McClinton and Sung, 1997). Therefore, it is likely that the unbranched trichome phenotype in fs mutants (Torres-Ruiz and Jürgens, 1994) is also linked to the microtubule phenotype. TON2 encodes a novel protein phosphatase 2A regulatory subunit, suggesting that microtubule organization in plants is controlled by the phosphorylation and dephosphorylation of proteins (Camilleri et al., 2002). Mutations in the SPIKE1 gene results in underbranched trichomes along with morphogenesis defects in various cell types. Microtubule organization is mis-regulated in spk1 mutants and the recent cloning of SPK1 revealed that it encodes an adapter protein involved in the integration of extracellular signals with the cytoskeletal organization (Qiu et al., 2002). These observations are supported by drug inhibitor studies that revealed distinct roles of actin and tubulin during trichome cell morphogenesis (Mathur et al., 1999; Szymanski et al., 1999; Mathur and Chua, 2000). Although the inhibition of the actin cytoskeleton causes irregularities in the directionality of cell expansion, drugs disturbing the microtubule cytoskeleton result in reduced trichome branching.

To further elucidate the molecular mechanisms underlying branch formation, we have studied the STI (STICHEL) gene. sti mutants exhibit the strongest branch phenotype: All trichomes are unbranched. Positional cloning revealed that STI encodes a novel protein containing a domain with sequence similarity to eubacterial DNA-polymerase III -subunits. However, endoreduplication levels were unaltered in sti mutants and in plants overexpressing STI. Also, the general organization of the actin and microtubule cytoskeleton in sti mutants was not affected. Our analysis of various sti alleles and transgenic lines overexpressing STI suggests that STI is a regulator of branch number rather than a requirement for branching.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Map-Based Cloning of the STICHEL Gene

We isolated the STI gene by molecular mapping of recombination breakpoints in about 1,800 F₁ events from a sti (Landsberg erecta [Ler]) × wild-type (Columbia-0 [Col]) cross and chromosomal walking. The STI gene was initially mapped north of the two molecular markers (m497 and m246) on chromosome II and further fine mapped to an interval of about 2 cM between ve012 and m246 as shown in Figure 1. This interval is spanned by two overlapping yeast (Saccharomyces cerevisiae) artificial chromosomes (YACs), CIC11A4 and CIC9D8. The right end of CIC9D8 and the molecular marker ve012 were used to establish a chromosomal walk with bacterial artificial chromosomes (BACs). Five additional CAPS markers were identified in this region that enabled us to map the STI gene to a 50-kb interval on the BAC F8N6 that lies within a region sequenced by the TIGR group of the Arabidopsis Genome Initiative (BACs T8K22 and T16F16). The annotation of the genomic sequence of this interval revealed seven potential genes (for details, see Fig. 1).

View larger version (19K): [in this window] [in a new window]   Figure 1.   Mapping of the STI gene. The genetic map between ve012 and m497 at the northern part of chromosome II is shown at the top. Below, Relative orientation of relevant YAC and BAC clones is displayed. BACs of the Institut für Genbiologische Forschung (IGF) library (F1E5, F5I7, and F8N6) were identified and mapped using probes marked as black squares. BACs of the TAMU library (T8K22 and T16F16, GenBank accession nos. AC004136 and AC005312, respectively) are from the mi320-nga1145 contig sequenced by The Institute for Genomic Research (TIGR; http://www.tigr.org/tdb/e2k1/ath1/) and were mapped based on sequence comparison with sequenced fragments of the IGF BACs. Molecular markers used for mapping are shown together with the recombination frequencies. The blowup of the chromosomal region between the two markers argyp and F8N6-sub#51 shows a higher resolution map and the predicted genes in the argyp-K22.12 interval, including the gene identification and name. Genes shaded in gray were sequenced for two or more sti alleles and the corresponding wild-type Ler.

To identify STI, we tested several candidate genes for allele-specific polymorphisms by sequencing genomic DNA of Ler wild type and two sti alleles. Both sti alleles have mutations in a gene containing a domain with sequence similarity to the eubacterial DNA-polymerase III -subunits (GenBank no. At2g02480). We confirmed that this gene represents the STI gene by identifying mutations in the genomic sequence of an additional five sti alleles. To determine the genomic structure of the STI gene, we used reverse transcription (RT)-PCR to amplify cDNA fragments from total RNA of young rosettes. Sequencing of the amplified cDNA fragments revealed a 3.6-kb open reading frame and the comparison with the genomic sequence showed that the STI gene contains six exons. The 3' end of the cDNA was determined by RACE-PCR. The transcription start could not be determined by RACE. However, we could amplify a cDNA fragment containing 120 bp upstream of the predicted START codon. This region contains one or more STOP codons in all possible frames, indicating that the predicted START codon initiating an open reading frame of 1,218 codons is correct. Figure 2 shows the STI cDNA sequence and the predicted STI protein sequence. The mutations found in sti alleles include small deletions and single-nucleotide exchanges that are predicted to result in premature STOP codons as well as mutations that are predicted to impair the correct splicing (for details, see Fig. 2). All changes are likely to eliminate the activity of mutant STI proteins. A sequence comparison between the two ecotypes (Ler and Col) revealed a number of nucleotide changes, 10 of which are predicted to cause amino acid changes (for details, see Fig. 2). In addition, we found a number of deletions and insertions ranging from 5 to 11 bp in some of the introns (data not shown), most of which are not predicted to affect splicing. Differential splicing was observed in intron 5. Some of the subcloned RT-PCR-amplified Ler-cDNA fragments were missing nine nucleotides at the beginning of the sixth exon, resulting from the utilization of an alternative splicing acceptor site (for details, see Fig. 2). As judged from the relative signal intensities when directly sequencing RT-PCR products, the two splicing events occur approximately with the same frequency (data not shown).

View larger version (58K): [in this window] [in a new window]   Figure 2.   Ler cDNA and predicted protein sequence of STI. STOP codons preceding the translation START are marked in bold italics. Intron positions are marked with >< above the nucleotide sequence. Intron 5 is differentially spliced in Ler, leading to the deletion of S K Q at the positions 1,182 through 1,184 (marked in bold italics) in approximately one-half of the transcripts. Other Ler-/Col-specific polymorphisms and mutations in various sti alleles (with their name in parentheses) are indicated above the nucleotide sequence (lowercase letters for nucleotide exchanges,  for deletions). Premature STOP codons resulting from mutations are marked in bold. sti-47, although a weak allele, has a point mutation resulting in the earliest observed STOP mutation at position 109. Possibly, an alternative START codon (position 484, marked in bold italics) is used that is in frame with the normal reading frame and provided with an acceptable Kozak sequence (Kozak, 1984). The weak sti-40 allele has a G to A exchange five nucleotides downstream from the splice donor site of the second intron, which presumably impairs normal splicing. The sti-XT1 allele has a one-nucleotide deletion that results in a frame shift at position 2,596, leading to a STOP three codons downstream. sti-56 has a G to A exchange at the splice acceptor site of the third intron, resulting in a frame shift and a STOP codon at position 2882. sti-EMU shows a one-nucleotide deletion resulting in a frame shift at position 3,070, leading to the altered amino acid sequence M L K L V A A W K A D E Q S R C K G STOP. The following amino acids are altered in Col: 145 (V to M), 277 (I to T), 355 (S to N), 439 (G to E), 826 (S to A), 828 (I to V), 889 (A to G), 891 (V to I), 1,023 (K to T), and 1,177 (H to Q). Conserved functional domains are underlined: presumptive PEST boxes (dotted), presumptive bipartite nuclear localization sequences (double underlines), and the DNA polymerase III -subunit homology domain (single underline).

Putative Functional Domains of the STI Gene Product

The STI gene encodes a protein of 1,218 amino acid residues with a predicted molecular mass of 135.3 kD. Sequence comparison with other known proteins and motifs identified three putative functional domains. A large domain between amino acids 454 and 799 shows sequence similarity to eubacterial DNA-polymerase III -subunits (Fig. 3A). The prokaryotic DNAPol III -subunit is the main component of the -complex, which is important for the formation of the replication initiation complex with the dimeric -subunit. In principle, the DNA polymerase III is able to perform DNA replication without the -subunit, but the processivity is lower by several orders of magnitude. Upon ATP binding, the -complex loads the -subunit onto a primer DNA template. Dissociation of the -complex from the -subunit to allow the polymerase to bind the -subunit requires ATP hydrolysis (Bertram et al., 1998). The similarity is 49% to 55% (identity 29%-35%) within the homology region. Similarity to the family of 36- to 40-kD ATP-binding subunits of replication factor C (RFC, also known as activator 1), the archaebacterial and eukaryotic functional counterparts of the bacterial  subunit (Chen et al., 1992a, 1992b), is less pronounced (42%-44% similarity, 23%-27% identity). Figure 3B shows the alignment of STI with this group of proteins, including the putative RFC from Arabidopsis. The latter is that member of a class of four putative RFCs in Arabidopsis that shows highest sequence similarity with the eukaryotic RFCs (Fig. 5). Figure 4 illustrates that STI and four STI homologs from Arabidopsis represent a phylogenetically separate branch, thereby defining a new, potentially plant-specific, subfamily among the -subunit homologs. The five proteins of this subfamily are markedly larger than the relatively small RFC proteins and the prokaryotic DNAPol III  subunits and show sequence similarity outside the RFC/ subunit domain (Fig. 5).

View larger version (91K): [in this window] [in a new window]   Figure 3.   Sequence alignment of STI with prokaryotic DNA polymerase III -subunits and the small subunits of eukaryotic and archaebacterial replication factor Cs. A, Alignment of the STICHEL sequence with related protein sequences of prokaryotic DNA polymerase III -subunits. Black-shaded amino acids are identical, dark gray-shaded amino acids are conserved, and light gray indicates weak similarity. The numbering at the top corresponds to amino acid positions in STI. A.a., Aquifex aeolius (aq  1855); B.s., Bacillus subtilis (Bsu0019); C.p., Chlamydias pneumoniae (CPn0040). B, Alignment of STI with related protein sequences of the small subunit of replication factor C. A.t., Arabidopsis (At1g2169); A.f., Archaeoglobus fulgidus (AF20608); M.t., Methanobacterium thermoautotrophicum (MTH241); S.c., yeast (YSCRFC2); H.s., human (Homo sapiens).

View larger version (90K): [in this window] [in a new window]   Figure 4.   Sequence comparison between STI with as its closest homologs. STI is a member of a class of five homologs that share sequence similarity outside the DNA polymerase III -subunit/RFC domain. Black-shaded amino acids are identical, dark gray-shaded amino acids are conserved, and light gray indicates weak similarity.

View larger version (22K): [in this window] [in a new window]   Figure 5.   Phylogenetic tree of the STI, the DNA polymerase III -subunit, and the RFC small subunit family. The phylogenetic tree of STI, the DNA polymerase III  subunit and the RFC small subunit family were calculated using only the core homology region corresponding to amino acids 449 to 799 in STI. The phylogenetic distance is shown as an unrooted dendrogram. The scale bar indicates 10% changes of amino acids. The bootstrap values are indicated for each branch. The five closest related STI homologs are bold and the four putative RFC-like proteins in Arabidopsis are underlined. The five closest STI homologs fall in a class that is separate from both, the prokaryotic DNA polymerase III -subunits and the eukaryotic RFC small subunits (the three groups are marked by circles). Note that only one of the Arabidopsis RFCs, At1g2169, is in the same group as the known eukaryotic RFC small subunits. A.a.=A. aeolius (aq  1855); B.s., B. subtilis (Bsu0019); C.p., C. pneumoniae (CPn0040); A.f., A. fulgidus (AF20608); M.t., M. thermoautotrophicum (MTH241); S.c., yeast (YSCRFC2); H.s., human.

Two regions, one between amino acids 273 and 304 and a second between amino acids 425 and 449, show similarity to PEST domains known to mediate rapid protein degradation. According to the score calculated based on the PEST hypothesis by Rodgers et al. (1986), the two PEST domains found in STI have a very high score of 9.24 and 9.58.

Three nuclear localization signals (NLSs) suggest that STI is targeted to the nucleus with a probability of 87%. One NLS is located in the N-terminal part, and the two others are located tandemly at the very C terminus.

STI Is a Regulator of Branch Number

The finding that STI contains a domain with sequence similarity to a putative DNA-polymerase III subunit suggests a functional link between the basic cell cycle/replication machinery and cell morphogenesis. To further elucidate the mode of STI function, seven different sti alleles and the genetic interactions of representative strong and weak alleles with other mutants affecting branching and/or endoreduplication were studied.

Although strong sti alleles invariably exhibit almost only unbranched trichomes, two weaker alleles have an increased frequency of two-branched trichomes (Table I). The weakest allele, sti-40, carries a mutation in the second intron 5 bp downstream from the splice donor site. Thus, it is likely that the splicing of the transcript is impaired (Pepper et al., 1994), only occasionally producing a correct messenger, whereas the wrong product leads to a premature STOP. This idea is supported by the fact that far stronger alleles have STOP codons nearer to the C terminus, thus allowing synthesis of a larger protein fragment that usually is associated with a less severe phenotype (Fig. 2). The second weak allele, sti-47, has a STOP after 36 amino acids and is, in fact, the most N-terminal STOP observed in sti alleles so far. A possible explanation is that in this allele translation might be reinitiated from the next available in-frame ATG with a minimal Kozak sequence (Kozak, 1984), which would result in a protein lacking the first 161 amino acid residues (Fig. 2). Thus, in both weak sti alleles it is likely that the gradual reduction of branching is caused by reduced levels of STI expression or activity. This suggests that STI regulates branch initiation in a dosage-dependent manner.

The genetic analysis of the strong sti-EMU allele has shown that the sti mutant phenotype can be partially rescued in double mutant combinations with nok but not with try (Folkers et al., 1997). Although nok and try mutant trichomes display a similar branch phenotype, the formation of up to five branch points, the two genes act in two different pathways. In try mutant trichomes, endoreduplication is increased 2-fold and, as a consequence, trichome cells are bigger and form more branches. In nok mutants, the production of extra branches is not coupled to the number of endoreduplication cycles. Our analysis of double and triple mutants with the weak sti-40 allele revealed that, in contrast to sti-EMU, the increase in the DNA content can rescue branching in the sti-40 background (Table II; Fig. 6). Although sti-40 trichomes have one branch point with a frequency of 78%, the double mutant with nok or try has mainly two branch points and the triple mutant sti-40 nok try has up to four branch points (Table II; Fig. 6). In the corresponding double mutant combinations with the stronger sti-EMU allele, a similar trend is observed (Folkers et al., 1997), though the rescue effect by nok is not enhanced by try (data not shown). It is important to also note that one of the strongest sti alleles, the x-ray-induced sti-XT1, is partially rescued in a nok mutant background (Fig. 6G). To test whether a reduction in the DNA content reduces trichome branching in the weak sti-40 mutant, sti-40 gl3 and sti-40 gl3 nok mutant combinations were generated. The decrease in endoreduplication in the gl3 background eliminates branching almost completely, and this effect can be rescued by additionally removing NOK in the triple mutant (Table II; Fig. 6I).

View larger version (124K): [in this window] [in a new window]   Figure 6.   Single and double mutant phenotypes of combinations with different sti alleles. Scanning electron micrographs of trichomes. A, wild type; B, sti-EMU; C, sti-40; D, double mutant sti-40 nok-122; E, double mutant try-EM1 sti-40; F, double mutant sti-40 gl3; G, double mutant sti-XT1 nok-122; H, triple mutant try-EM1 sti-40 nok-122; I, triple mutant sti-40 nok-122 gl3; J, 35S:STI rosette leaf with four branched trichomes; K, 35S:STI leaf trichome with seven branch points; L, 35S:STI stem trichomes with branches.

The analysis of sti-alleles and their function in various mutant backgrounds suggest that STI acts in a dosage-dependent manner. To test this further, we transformed 35S:STI into sti-mutants. Sixty-seven transgenic lines were studied for the rescue of the mutant phenotype. Seventeen lines showed a weak rescue, and 34 showed complete rescue with up to three branches. In addition, we found in 11 lines trichomes with more than two branch points, occasionally up to five branch points (Fig. 6, J and K). Semiquantitative RT-PCR analysis of the 35S:STI-specific expression levels did not reveal clear differences in the RNA levels between lines exhibiting weaker or stronger rescue (Table III; Fig. 7A). Thus, in summary, the overexpression of STI does not only rescue the sti phenotype, but also leads to extra branch formation.

View larger version (45K): [in this window] [in a new window]   Figure 7.   STI expression. A, Analysis of expression levels in 35S:STI lines by semiquantitative RT-PCR. The numbers of the lines correspond to those in Table III. The elongation factor1 EF1 was used as a control. B, Northern-blot analysis of STI expression in different plant organs: lane 1, roots; lane 2, stems; lane 3, siliques; and lanes 4 and 5, rosette leaves. 28S rRNA served as load control for the blot.

Also, in 35S:STI lines, stem trichomes that are normally unbranched exhibited up to two branch points (Fig. 6L), suggesting that organ-specific differences in trichome branching are controlled by STI.

STI Is Expressed in All Organs

To test where STI is expressed, we performed a northern-blot analysis. STI expression was found at about the same levels in all organs tested: rosette leaves, siliques, stem, and roots (Fig. 7B). This finding was confirmed by RT-PCR using primer pairs that specifically amplify STI but not its homologs (data not shown). Also, rosette leaves from gl1 (glabra1) mutants that are devoid of trichomes show STI expression at approximately the same levels as wild type. These data show that STI is expressed ubiquitously in all organs rather than in a trichome-specific manner.

STI Is Not Involved in the Regulation of Endoreduplication

As judged by the visual inspection of 4',6-diamino-phenylindole-stained whole mounts, sti trichomes appear to have a DNA content similar to wild type. To exclude that reduced branching in sti mutants is due to a reduction in the DNA content, we measured the DNA content in trichomes of the weak sti-40 allele and the strong sti-XT1 allele. In both alleles, it was on average 37.5 ± 18 C (n = 50) and 41.4 ± 15 C (n = 58), respectively, which closely corresponds to wild-type measurements (38 ± 17 C, n = 49). Also, sti try double mutants showed the expected 2-fold increase in the DNA content (sti-40 try, 93 ± 35 C, n = 51; sti-XT1 try, 84 ± 33 C, n = 57; and try, 80 ± 37 C, n = 49), again indicating that DNA replication in sti trichomes is not impaired.

Conversely, overbranched trichomes in 35S:STI plants did not have an increased DNA content. This was studied in two independent lines. In both lines, trichomes had a DNA content indistinguishable from wild type (38 ± 15 C, n = 149; 37 ± 16 C, n = 60).

STI Does Not Affect the Actin or Microtubule Organization

In trichomes, actin and tubulin appear to have distinct roles such that during the early growth phase until branching is completed, microtubules play a major role, whereas actin is important during cell expansion (Mathur et al., 1999; Szymanski et al., 1999; Mathur and Chua, 2000). To study whether the actin or the microtubule cytoskeleton are affected in sti mutants, we compared their arrangement in wild type and sti mutants.

To monitor the actin cytoskeleton, we used a transgenic 35S::GFP-talin line. This line ubiquitously expresses the green fluorescent protein (GFP) fused to the C terminus of mouse talin, which binds to F actin and, thus, decorates the actin in vivo (Rees et al., 1990; McCann and Craig, 1997; Kost et al., 1998). In mature wild-type trichomes, the cortex of the cell is characterized by a fine actin network. The more prominent bundles in the center of the cell are mainly oriented with respect to the longitudinal axis. In sti mutants, the actin distribution was indistinguishable from wild type, indicating that STI is not involved in the regulation of the actin cytoskeleton (data not shown).

To visualize the microtubule distribution in living trichomes, we used a transgenic 35S::GFP-MAP4 line that expresses GFP fused to the microtubule-associated protein 4 (Mathur and Chua, 2000). In mature wild-type trichomes, the microtubules are found mainly at the cell cortex where microtubule filaments are initially radially and later, after branch formation is completed, longitudinally oriented (Fig. 8). Also, sti mutants showed a shift from an initial radial microtubule organization to a longitudinal orientation.

View larger version (189K): [in this window] [in a new window]   Figure 8.   Microtubule organization in wild-type and sti mutant trichomes. A, Microtubule organization at the base of a wild-type trichome. B, Microtubule organization at the branch point of a wild-type trichome. C, Microtubule organization at the base of a sti trichome. D, Microtubule organization in the mid-region of an sti mutant trichome.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Although the genetic and cell biological analysis of trichome branching provides a well-defined framework for the formal logic of the system, little is known about the underlying molecular mechanisms. The molecular analysis of several branching mutants revealed links to the control of the microtubule function at different regulation levels, including the microtubule-based transport processes (Oppenheimer et al., 1997), the regulation of microtubule assembly and disassembly (Burk et al., 2001; Webb et al., 2002), and the control of microtubule organization (Traas et al., 1995; Camilleri et al., 2002; Folkers et al., 2002; Kim et al., 2002; Qiu et al., 2002). Although these findings provide an excellent entry point into the understanding of the final steps of cell morphogenesis, earlier steps such as the control of branch initiation and its spatial control remain misunderstood. Because genetic analysis has suggested that STI plays a key role in these early processes we have isolated the STI gene by a positional cloning approach. The molecular characterization of seven sti alleles revealed allele-specific polymorphisms that are all predicted to severely impair either splicing and/or protein function. Therefore, we conclude that we have cloned the STI gene.

The Role of STI in Cell Morphogenesis

Three lines of evidence suggest that STI regulates branching in a dosage-dependent manner. First, mutations in the STI gene do not simply eliminate its function, but depending on the severity of the defect, intermediate phenotypic defects are also observed. This suggests that less STI activity results in fewer branches. Second, conversely, lines overexpressing STI can trigger extra branch formation. A third line of evidence supporting a regulatory role of STI in branch formation comes from the genetic analysis of double mutants. In a previous study, the finding that sti mutants can be rescued in double mutants with nok but not with try has led to the assumption that STI and NOK might specifically counteract each other (Folkers et al., 1997). However, our findings that weak sti alleles can also be rescued by try and that the additional removal of TRY in a sti nok background results in an even better rescue suggests that mutations in the sti gene can be bypassed in several ways. This suggests that STI is not required to make branches, but involved in the regulation of their number.

The cell biological analysis of sti mutants revealed no deviation from wild type at the subcellular level. One criterion to monitor cell differentiation is the timing and extent of cell vacuolization. Reduced vacuolization was found to be associated with severe growth abnormalities of root hairs in rhd3 mutants (Galway et al., 1997). Vacuolization in sti mutant trichomes, however, was normal (data not shown). A second important aspect is the organization and function of the actin and microtubule cytoskeleton. The general organization of both cytoskeletal elements was normal, suggesting that STI is not involved in the control of the microtubule or actin organization.

Potential Molecular Function of STI

The sequence similarities of STI to other proteins provides few clues about its molecular function. The presence of NLS domains and a DNA-polymerase III -subunit/RFC domain suggest that STI might be involved in the regulation of DNA replication. This, however, seems to be unlikely because the ploidy level in trichomes is normal in sti mutants and in lines overexpressing STI, indicating that replication is not affected in both situations. Consistent with this interpretation is the finding that STI belongs to a group of five genes that is clearly distinct from the putative Arabidopsis RFC genes. These five genes also show sequence similarity outside the DNA-polymerase III -subunit/RFC domain, suggesting that they may have adopted a new plant-specific role. If STI, as suggested by the NLS motives, is localized to the nucleus, STI might be involved in the regulation of gene expression. However, because it lacks any similarity to known transcription factors, this would be by a novel mechanism. An attractive alternative would be that STI is involved directly in the formation of branches in the cytoplasm. In analogy to the role of DNA-polymerase III -subunit/RFC, STI could mediate the formation of protein complexes initiating branch formation. Localization studies should allow the first insights in these questions.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Plant Material and Genetic Analysis

Plants were grown at constant illumination at 23°C. Wild-type strains used in this work were Ler and Col. Single mutant lines were isolated from the F₂ progeny of plants mutagenized with ethyl methanesulfonate as described (Mayer et al., 1991) with the exception of the sti allele sti-Cs430, which was isolated from a T-DNA insertion mutagenesis screen by Csaba Koncz (Max-Planck-Institut, Cologne, Germany). Double mutants were created by preselecting F₂ plants displaying the phenotype of one of the parents and selecting F₃ plants for a new phenotype. Double mutants were backcrossed with the parental lines to verify their genotype. Triple mutants were created from double mutants that shared one of the mutant backgrounds following the same procedure. The following single and double mutants have been described previously: sti-EMU, sti-EM1, sti-XT1, try-EM1, and try-EM1 sti-EMU (Hülskamp et al., 1994); sti-40, nok-122, nok-122 sti-EMU, and nok-122 sti-40 (Folkers et al., 1997); and gl3 (Koornneef et al., 1982).

Molecular Cloning of the STI Gene

Plant genomic DNA was prepared as described previously (Leutwiler et al., 1984; Lukowitz et al., 1996). For initial mapping, the RFLP marker probes m246 and m497 modified by Fabri and Schäffner were used (Fabri and Schäffner, 1994). m246 was further used as a CAPS marker (Konieczny and Ausubel, 1993). The YAC contig was established by Zachgo et al. (1996; http://weeds.mgh.harvard.edu/goodman/). The hybridization probes for the chromosome walk were generated by modified thermal asymmetric interlaced-PCR (Liu and Whittier, 1995) with the nested primers 5'-CAG CAA CCG CAC CTG TGG C-3', 5'-GAT GCG TCC GGC GTA GAG-3', and 5'-GAG TCG AAC GCC CGA TC-3' for the right end of CIC9D8 and 5'-GAT TAA GTT GGG TAA CGC CAG GG-3', 5'-AGT CAC GAC GTT GTA AAA CGA CG-3', and 5'-GTA ATA CGA CTC ACT ATA GGG C-3' for the BAC T₇ ends. The BAC clones from the library IGF-BAC 122 (Mozo et al., 1998) were received from Katharina Henschel (Resource Center/Primary Database of the German Human Genome Project, Berlin).

New CAPS marker were generated by the following strategy: BACs were shotgun subcloned into pBluescript SK⁺ (Stratagene, La Jolla, CA). Inserts of appropriate size were sequenced and primers for genomic amplification were designed with the McVector 4.5 software (Kodak Scientific Imaging Systems, New Haven, CT). PCR products were tested for ecotype-dependent restriction enzyme polymorphisms. To convert the RFLP marker mi320 (Robert Whittier, Mitsui Plant Biotechnology, Tsukuba, Japan) into a CAPS marker, the 1.3-kb RFLP marker probe (received from the Arabidopsis Biological Resource Center, Columbus, OH, stock no. CD3-281) was used to identify a 6.4-kb HindIII fragment from BAC F20I2. An ecotype-specific polymorphism was identified as described above. Sequence information on ve012 was provided by David Bouchez (Laboratoire de Biologie Cellulaire, Institut National de la Recherche Agronomique, Versailles, France). K22.12, K22.20, argyp, and nalf were designed based on sequence information submitted by the TIGR group of the Arabidopsis Genome Project (see Fig. 1). The CAPS marker primers were: ve012, 5'-CCC ACC ATT GAA GGA GAA AAC AG-3' and 5'-CAG CAG AGA AGA AGT AAA CCG CC-3' (Eco130I); mi320, 5'-CTG TCT TTT GGT ATT TCG GG-3' and 5'-GAA CGA TCA ATG TGT TGG AAG-3' (HindIII); argyp, 5'-GTT GCT CGG CAG AAG TTA G-3' and 5'-AGG TAC TCG CAC AGT ACG C-3' (Csp6I); nalf, 5'-GCT GAA ATA TCA TTG TGC C-3' and 5'-GCC CAT TAA AGG ATA GTG G-3' (length polymorphism); K22.12, 5'-CTA TAG CAA ACT CAT GGC-3' and 5'-AGT CTA AGA TGA TGG TGG-3' (MvaI); K22.20, 5'-CAT GAA ATT CCC TTC CAC-3' and 5'-ATC GTT ACC AAA GCA TCC-3' (AvaI); and F8N6-sub#51, 5'-AGT CTC CCT TAG TTT GAG TG-3' and 5'-CTG GAA TAG AAG CAT TAC C-3' (HpaII). Further details are available on request.

Primers for amplification and sequencing of gene candidates were also created based on the sequence information provided by TIGR. Sequencing was carried out on an ABI Prism (Perkin-Elmer Applied Biosystems, Foster City, CA) sequencing equipment according to the manufacturer's instructions. The STI-transcribed sequence was amplified using the 5'-TGA AAA CGC GAA GCT GAG AGA G-3' and 5'-CGA ACA GGA GTT CCC TTG-3' primer pair from cDNA obtained by oligo(dT)-primed RT of RNA isolated from rosettes with four to five leaves. The 3' end was obtained using the oligo(dT) primer and the gene-specific nested primers 5'-CTG ATA AAG ACA CAC CTG GAT CG-3' and 5'-ATC GCC AAA CTA ACG TAG C-3'. 5'-RACE PCR was performed with the 5'-RACE System for Rapid Amplification of cDNA Ends, version 2.0 (Life Technologies/Gibco-BRL, Cleveland) according to the manufacturer's instructions using the specific nested primers 5'-CAT TAA CAC TAG CTT GCG TCC AC-3' and 5'-CTC TTC TTC CTT ACC ATT CTT AG-3'. It yielded no products related to the STI gene. The sequence of the 5' region, including 120 bp of upstream untranslated sequence, was obtained using the 5'-TTG CAC AGG TTT TGA AAT GTC AG-3' and 5'-CTC TTC TTC CTT ACC ATT CTT AG-3' primer pair.

The expression levels of STI-overexpressing plants were estimated by semiquantitative RT-PCR analysis. RNA was isolated using Dynabeeds (Dynal Biotech, Oslo) and treated with DNaseI to remove genomic DNA. RT PCR was carried out with Titan One tube RT-PCR mix (Roche Diagnostics, Indianapolis). The primer pair SNT-70-forward (5'-CGA CGG TAT CGA TAA GCT TG-3') and SNT-70-reverse (5'-ACA CCT AAA ACC ACC GAA G-3') were designed to only amplify the 35S:STI transcript. No transcript was detected in wild-type plants after 35 cycles.

Sequence Analysis

Proteins related to STICHEL were identified by BLAST search (Altschul et al., 1990). Sequence alignment was performed with the Multiple Sequence Alignment Program (Huang, 1994) at the Baylor College of Medicine Search Launcher (http://searchlauncher.bcm.tmc.edu/multi-align/), BOXSHADE version 3.21 by K. Hoffmann and M. Baron at EmbNet (ftp.isrec.isb-sib.ch), and Vector NTI alignX-Block version 7.0. Functional domains were identified by searching the protein sequence against the Pfam database (Bateman et al., 1999) and PSORT 6.4 (http://psort.nibb.ac.jp). The phylogenetic tree was created using distance, parsimony, and maximum likelihood criteria using the PAUP 4.0B.1 program (Swofford, 1998). The confidence of branching was determined using 1,000 bootstrap resamplings.

Cytological Methods and Pictures

Confocal laser-scanning microscopy was done with the TCS-NT program (Leica Microsystems, Bensheim, Germany). 4',6-Diamino-phenylindole staining and nDNA content measurements were performed as described previously (Hülskamp et al., 1994). For scanning electron micrographs freshly cut leaves were mounted without fixing and immediately analyzed at 10 kV. Pictures were processed with the Aldus Freehand 7.0 (Aldus Corp., Seattle) and Adobe Photoshop 3.0 (Adobe Systems Inc., Mountain View, CA) software.