The Arabidopsis gene MONOPTEROS encodes a transcription factor mediating embryo axis formation and vascular development



The vascular tissues of flowering plants form networks of interconnected cells throughout the plant body. The molecular mechanisms directing the routes of vascular strands and ensuring tissue continuity within the vascular system are not known, but are likely to depend on general cues directing plant cell orientation along the apical–basal axis. Mutations in the Arabidopsis gene MONOPTEROS (MP) interfere with the formation of vascular strands at all stages and also with the initiation of the body axis in the early embryo. Here we report the isolation of the MP gene by positional cloning. The predicted protein product contains functional nuclear localization sequences and a DNA binding domain highly similar to a domain shown to bind to control elements of auxin inducible promoters. During embryogenesis, as well as organ development, MP is initially expressed in broad domains that become gradually confined towards the vascular tissues. These observations suggest that the MP gene has an early function in the establishment of vascular and body patterns in embryonic and post-embryonic development.


Vascular tissues form ramified systems of continuous cell files, each made of elongated, interconnected cells (Steeves and Sussex, 1989; Lyndon, 1990). Vascular patterning is thought to depend on signals directing the routes of vascular strands as well as the oriented differentiation of each cell within the vascular system. The molecular nature of these signals is not known, but (canalized) flows of translocatable signal molecules, possibly involving the plant hormone auxin, have been implicated in vascular differentiation (reviewed in Shininger, 1979; Lyndon, 1990; Sachs, 1991; Nelson and Dengler, 1997).

Embryonic and post-embryonic development are tightly linked to the patterning of the internal provascular systems and are therefore thought to depend, at least in part, on common directional signals (Sachs, 1991; Cooke et al., 1993). Although vascular tissues differentiate at predictable positions during normal development, adaptive responses to wounding or abnormal growth conditions demonstrate considerable flexibility of vascular patterning and enable investigation of the nature of the underlying signals. Local application of the plant hormone auxin has been shown to influence efficiently the vascular pattern in mature organs (Sachs, 1981). When applied at early stages, auxin, as well as chemical inhibitors of auxin transport, severely affects the architecture of the embryo (Liu et al., 1993; Fischer and Neuhaus, 1996). These observations suggest an involvement of the plant hormone auxin in mechanisms co-ordinating the development of embryos and the respective vascular patterns.

A number of mutants affecting various aspects of the vascular pattern have been identified (reviewed in Freeling, 1992; Nelson and Dengler, 1997), but only two mutants have been described to affect vascular tissue continuity throughout the plant body. In the Arabidopsis mutant lopped 1 (Carland and McHale, 1996), interruptions of vascular strands are associated with abnormal spiral growth of all organs and with gross abnormalities in leaf shape. Mutations in the gene MONOPTEROS (MP) interfere with the formation of the vascular system already in the embryo (Berleth and Jürgens, 1993). Early in embryogenesis, mp mutants lack centrally located provascular cells within a basal domain of the embryo. This domain gives rise to the hypocotyl (seedling stem) and the primary root (collectively referred to as the embryo axis; Figure 1A and C) (Berleth and Jürgens, 1993; Przemeck et al., 1996). Both structures are completely missing in mutant embryos and seedlings (Figure 1B and D) and the mutant has therefore been classified as a potential embryo pattern mutant (Mayer et al., 1991). Occasionally, mp mutant seedlings can produce adventitious roots enabling studies of mutant traits at post-embryonic stages. In all organs analyzed, cells within vascular strands appear incompletely differentiated and insufficiently interconnected (Przemeck et al., 1996), and in all leaf organs the vascular system is reduced to higher order veins (for example, see Figure 1F and G). Furthermore, there are variable distortions in the formation of lateral organs, particularly in the inflorescence (Przemeck et al., 1996). This abnormality is reminiscent of plants treated with chemical inhibitors of auxin transport (Okada et al., 1991). Auxin transport is reduced in mp mutants, even when measured in stem segments of mutants from weak alleles that do not display marked vascular abnormalities (Przemeck et al., 1996). Based on these observations, the MP gene has been proposed to mediate plant cell axialization in response to directional cues along the apical–basal axis (Przemeck et al., 1996).

Figure 1.

(A) Wild-type embryos at the onset of heart-stage (cotyledons just emerged) display files of elongated cells in all subepidermal tissues of the incipient hypocotyl/root region (separated from apical region by an arrow in A). Arrowhead in (A) points at elongated cells of the incipient vascular system; ep, epidermis; su, suspensor. (B) Cells in the corresponding part of mp mutant embryos are neither elongated nor organized in files, while unrelated structures, such as the suspensor and the epidermis, are not affected. (C) The wild-type seedling consists of a shoot meristem (not visible), two cotyledons, the hypocotyl and the primary root. (D) In mp mutants cotyledons may be variably fused and the hypocotyl/root axis is replaced by a basal peg of unorganized tissue. (E)The vascular system in wild-type cotyledons comprises lobes extending from a central vascular strand. (F) Vascularization is reduced to central strands with small terminal branches in mutants of the weak allele CSH1. (G) Vascularization is further reduced to central veins or less in individuals of a strong allele, U252. Allele-specific differences are also reflected in a number of other defects, such as integrity of individual vascular strands and the frequency of cotyledon fusions (data not shown).

In this report we describe the isolation and molecular characterization of the MP gene. We show that the MP gene encodes a protein with features of a transcriptional regulator that is very likely to be capable of modulating gene activities in response to auxin signals. These molecular properties, as well as the MP expression profile, are consistent with the mutant phenotype and suggest that MP influences embryo pattern formation as well as vascular development by mediating axialized behavior of plant cells in response to auxin cues.


Isolation of the MP gene

The MP locus has been mapped to the upper part of the first chromosome (Berleth and Jürgens, 1993). Using restriction fragment length polymorphism markers (RFLP) from this region, we have localized the gene between markers m59 and g2395 (Hardtke and Berleth, 1996). Within this region, ∼2200 kbp of contiguous genomic DNA from libraries of Yeast Artificial Chromosome (YAC) clones were isolated (Figure 2A). Mapping of RFLPs detected by DNA fragments from the chromosome walk enabled us to assign the MP gene to a single YAC that was utilized to identify and map new RFLP loci in the immediate vicinity of the MP locus. Based on these new RFLP loci, a local chromosome walk of cosmid and Bacterial Artificial Chromosome (BAC) clones was generated, encompassing a genetic interval defined by two recombination events on either side of the MP gene (Figure 2A). Eight classes of non-overlapping cDNA clones were identified within this interval.

Figure 2.

(A) Map of the MP genomic region with adjacent RFLP markers (top; Chang et al., 1988; Nam et al., 1989), approximate genetic distances and the extent of the chromosome walk (shaded rectangle; Hardtke and Berleth, 1996). The approximate positions of YAC clone CIC8C5 (Creusot et al., 1995) (∼450 kbp, open rectangle) within the chromosome walk and of the aligned cosmid (thin lines) and BAC (bold line) clones within CIC8C5 are indicated below. Numbers give genetic distances in cM (top) and the fraction of recombinations in chromosomes from a total of 1796 meiotic products (bottom). The aligned cosmid and BAC clones span a region of roughly 60 kbp. (B) Southern blot analysis. DNA from plants heterozygous (het) or homozygous (hom) for the indicated mp mutant alleles hybridized with 32P-labeled MP cDNA clone KS 18 (1.8 kbp). The three alleles, G33, G25 and G12, were induced in Col-0 by gamma-ray mutagenesis. The single band in G33 and G25 represents the absence of detectable restriction site alterations (for G33 consistent with the single base pair deletion shown in Figure 2C). Arrowheads mark the positions of the bands affected by allele specific restriction site polymorphism in allele G12. The molecular lesion in the gamma-ray-induced allele G12 was not further characterized, as it appears to represent a larger chromosomal defect associated with multiple RFLPs (data not shown). (C) Structure of the MP transcription unit. The diagram depicts introns (thin lines) and exons (5′ and 3′ untranslated regions, bold lines; ORF, shaded boxes). Sizes are given in nucleotides. The only long open reading frame is preceded by six in-frame stop codons. Arrows below mark the positions of sequence alterations in six mp alleles. Codon positions refer to the predicted protein product shown in (D). (D) Predicted amino acid sequence (single-letter code) of the MP protein product (Col wild-type). Numbers to the right indicate the position relative to the putative translational start site. Positions of the potential functional domains are marked by rectangles on top: DNA-binding domain (shaded), nuclear localization sequences (filled, domains of bipartite NLSs connected by thin line), a potential helix–loop–helix domain predicted by structural features (Rost and Sander, 1994) (hatched) and two stretches of similarity to ARF1 (Ulmasov et al., 1997) and to a number of auxin inducible genes (Abel et al., 1995) (open). (E) Amino acid sequence comparison (single-letter code). An alignment is shown for the deduced amino acid sequences from the gene products of VP1 (McCarty et al., 1991), MP (shaded rectangle in D) and ARF1 (Ulmasov et al., 1997). Identical and functionally conserved amino acids are indicated by # and *, respectively.

To detect allele-specific RFLPs, cDNA clones representing the eight presumed transcription units in the region were hybridized to genomic Southern blots of mutant DNA. Two overlapping cDNA clones (KL1 and KS18) identified an allele-specific DNA polymorphism (Figure 2B). The longer cDNA clone, KL1 (2.7 kbp) was extended by RACE–PCR to obtain sequence information for 3.1 kbp of transcribed DNA. Both cDNAs detected a single low-abundance 3.2 kbp transcript on poly(A)+ RNA blots (data not shown) suggesting that the cDNA sequence represents the full-length transcript. Comparison of the cDNA and corresponding genomic sequences indicated a transcription unit of 13 exons spread over a genomic interval of ∼4.5 kbp (Figure 2C). We confirmed that this transcription unit represents the MP gene by analyzing the genomic sequences from six mp alleles. Direct sequencing of PCR products revealed that all six alleles had either stop or frame-shift mutations at different positions within the predicted coding sequence (Figure 2C). Based on the perfect cosegregation and the existence of deleterious point mutations in six mutant alleles, we conclude that the isolated transcription unit represents the MP gene.

Sequence analysis and nuclear localization of the presumptive MP product

The open reading frame of the MP gene encodes a predicted protein product of 902 amino acids and contains three stretches of similarity with the recently described Arabidopsis transcription factor ARF1 (Ulmasov et al., 1997). Sequence similarities between MP and ARF1 are particularly pronounced within a presumptive DNA binding domain (76% similarity between residues 150–264, Figure 2E) that is related to Maize transactivator Viviparous 1 (McCarty et al., 1991) (Figure 2E). The MP sequence is most probably partially represented by cDNA IAA24 (Ulmasov et al., 1997) (865 out of 866 amino acids identical). The DNA binding domains of ARF1 and IAA24 have recently been shown to bind to the same functionally defined promoter elements of auxin-inducible genes (Ulmasov et al., 1997), suggesting that these proteins regulate downstream genes in response to auxin signals. In the C-terminal region, the products of MP, ARF1 and a larger group of otherwise unrelated auxin-inducible genes (Abel et al., 1995), contain two additional stretches of similarity. These two regions (residues 794–827 and 838–880, Figure 2D) (Ulmasov et al., 1997) have (for ARF1) been implicated in protein–protein interaction.

Interestingly, all six mp mutations characterized appear not to affect the DNA-binding domain (Figure 2C). Since all available mp mutants have been identified at the seedling stage, this raises the possibility that true null-alleles are missing due to early embryonic lethality. All mp alleles analyzed delete the C-terminal stretches of homology and could thus reflect the requirement of protein interactions for full gene activity. Furthermore, premature stop codons at different positions in the central portion of the predicted MP protein are associated with two distinguishable phenotypes. Mutants of alleles CSH1 and G92 (premature stop codons at residues 630 and 657, respectively; Figure 2C) display intermediate vascular defects (Figure 1F), while stronger phenotypes are observed in alleles with stop codons (or frame-shift mutations) at earlier positions (Figures 1G and 2C). Distinguishable weak and strong phenotypes have also been reported for previously isolated mp alleles (Berleth and Jürgens, 1993).

The predicted MP product contains three potential nuclear localization signals (NLSs) within the presumed DNA-binding domain (Figure 2D). One of these NLSs is of bipartite structure, while two show similarity with the MATα-class of NLS (Raikhel, 1992). Two further potential NLSs were identified outside the DNA-binding domain (Figure 2D). To test the ability of the MP product to exert a nuclear function and to delimit sequences essential for nuclear import, we have assessed the capacity of MP protein domains to confer nuclear localization to an attached β-glucuronidase (GUS) reporter gene product by assaying transient expression in onion epidermis cells. A blue precipitate monitoring GUS activity was confined to the nuclei in cells transformed with an MP–GUS fusion construct, containing residues 4–289 of the predicted ORF, indicating the functionality of at least one of three proposed NLSs within the DNA-binding domain (Figures 2D and 3A). By contrast, a second MP–GUS fusion protein containing residues 297–901 of the predicted MP product was not selectively imported into the nucleus (Figures 2D and 3B). We conclude that functional NLSs are located in the N-terminal part of the protein and that nuclear import is not dependent on possible molecular interactions of the C-terminal domains that are missing in mutant gene products.

Figure 3.

Expression of MP–GUS fusion proteins in onion epidermis cells. (A) Expression of a fusion protein containing the presumed DNA-binding domain with three potential NLSs (residues 4–289) of the predicted MP protein fused to the N-terminus of the GUS gene. (B) Expression of a similar construct as in (A), comprised of most of the remaining coding sequence (residues 297–901) of the predicted MP product including two more NLS-related sequences that may also constitute another bipartite NLS. (C) Control of expression of the GUS gene in the same vector.

Expression of the MP gene in embryonic and post-embryonic development

In RNA blot analyses, low levels of MP transcripts were detected in all major organs (data not shown) consistent with ubiquitous vascular distortions (Przemeck et al., 1996). In order to better correlate the expression pattern to the phenotypic defects, we determined MP transcript distribution by in situ hybridization to tissue sections of embryos and plant organs. In early globular embryos, MP transcripts were present in all subepidermal cells (Figure 4A), while in heart-stage embryos expression was confined to broad, yet more central domains along the midlines of the cotyledons, as well as of the embryo axis (Figure 4B). Expression was further restricted to the centers of embryonic organs in early torpedo-stage embryos (Figure 4C) and was ultimately confined to provascular tissues of the differentiating vascular strands in nearly mature embryos (Figure 4D). Similarly, MP was expressed in broad domains in emerging determinant shoot organs (Figure 4F and H), while in more mature organs expression was restricted to procambial and possibly to some differentiated vascular regions (Figure 4F, G, L and M). In early leaf primordia, ubiquitous (yet subepidermal) expression was observed that became gradually restricted to vascular tissues upon leaf maturation (Figure 4G). In early flower primordia, MP was expressed in all whorls (Figure 4H), but was confined to vascular tissues in maturing flower organs. At late stages of flower development, expression was most pronounced in the gynoecium, particularly in developing ovules including the funiculi (Figure 4K). In mature roots, MP expression was detectable only within the central cylinder (Figure 4I).

Figure 4.

Expression pattern of MP mRNA in wild-type (Col-0) embryos and plants. In situ hybridization with MP sense (E) and antisense (A–D, F–M) probe. (A–D) Progressive central confinement of MP expression in embryos from early-globular (A) through early-heart (B) and early-torpedo (C) stages (median longitudinal sections in A–C). In the nearly mature embryo [cross-section in (D)] MP expression is restricted to provascular tissues in the cotyledons and in the hypocotyl. (F) Median longitudinal section of shoot apical meristem of a 7-day-old seedling showing broad subepidermal MP expression domains in young leaf primordia (arrow) and predominantly procambial expression in more mature leaves (arrowheads). (H) Longitudinal section of an inflorescence apex. Broad, strong MP expression in flower and floral organ primordia of different stages and weaker procambial expression (arrowheads). No detectable expression in the inflorescence meristem apex. (K) Gynoecium displaying MP expression in immature ovules and vascular tissues (arrowhead). (G–M) MP expression confined to the central root cylinder (I) and to vascular bundles in a mature cauline leaf (G) and inflorescence stem [cross-section in (L), longitudinal section in (M)]. Abbreviations: cot, cotyledons; hyp, hypocotyl. Scale bars: 50 μM in A-E, H, K-M; 200 μM in F, G, I.


In this study, we have determined and analyzed the molecular identity and expression profile of the MP gene product. The MP gene had initially been identified by mutations that severely distort the embryonic pattern (Berleth and Jürgens, 1993) and subsequent studies suggested that the embryonic defects reflect a more general incapacity of mutant cells to respond to apical–basal axial cues that are instrumental for both embryo axis formation and vascular development (Przemeck et al., 1996). Here we show that the MP gene encodes a protein partially represented by a recently identified transcriptional regulator (Ulmasov et al., 1997) whose binding properties match the expectations of widely accepted concepts of plant apical–basal signaling (Sachs, 1991). We further show that MP expression in early embryo and organ development coinincides with axial cell orientation and later converges towards the routes of vascular differentiation, consistent with an early function in vascular strand formation.

The role of the MP gene in vascular development

The molecular mechanisms underlying vascular patterning and differentiation are poorly understood. Genetic approaches in Arabidopsis have been hampered by the difficulty of visualizing the vascular system in mutant screens and only a very small number of mutants have been isolated, either based on anatomical defects (Turner and Somerville, 1997) or by associated morphological abnormalities (Berleth and Jürgens, 1993; Carland and McHale, 1996). The Arabidopsis mutant mp was initially recognized by its conspicuous seedling phenotype lacking all structures derived from the basal domain of the embryo (Mayer et al., 1991). This localized defect suggested a region-specific organizing function of the MP gene during plant embryo pattern formation (Mayer et al., 1991; Berleth and Jürgens, 1993). When analyzed at the anatomical level, however, mp mutant organs of all developmental stages display a unique type of vascular defect characterized by an overall reduction of vascular tissues and by incomplete tissue continuity within vascular strands (Przemeck et al., 1996). No general cellular defects were detected, and even within the vascular system all classes of differentiated cells can be observed (Przemeck et al., 1996). The mp gene function thus seems to be required to mediate the integrated formation of vascular cell files rather than to promote particular events during vascular differentiation. In this study we show that MP is very specifically expressed in the vascular tissues at all stages of vascular maturation. Unlike other genes implicated in vascular development, however, this tissue-specific signal is preceded by an extremely early expression in far broader domains that become gradually confined towards the sites of vascular differentiation. These observations suggest an early function of the MP gene in organ initiation and the gradual sharpening of the MP expression domains seem to reflect genetic interactions establishing the vascular pattern during organogenesis.

Axial cues in vascular and organ development

The distribution of MP transcripts in developing embryos does not support the earlier concept of a complex region-specific organizing role of the gene in the basal domain of the embryo. Rather, the molecular data support the view that the basal focus of the mutant embryonic phenotype results from a localized requirement for axial information at the onset of hypocotyl/root axis formation in the early embryo. Vascular differentiation and the generation of the hypocotyl/root body region appear to be developmentally linked and may thus be related at the cellular level. In fact, the general concept of an apical–basal signal flux underlying oriented cell behavior has not been restricted to vascular strand formation, but has also been applied to cell orientation in primordia (Sachs, 1991; Cooke et al., 1993). Assuming that the initiation of axiality is more sensitive than its maintenance, it seems plausible that the generation of continuous cell files from previously isodiametric cells, as it occurs in the early embryo or during lower-order vein initiation in leaf blades, is particularly dependent on proper perception of axial signals. These structures are most severely affected in mp mutants.

Irrespective of the molecular nature of these signals, the mp phenotype in alleles of different strength demonstrates that vascular and embryonic pattern formation are correlated. Mutants carrying weak mp alleles display a spectrum of seedling phenotypes, including those with short stretches of basal vascular tissue (Berleth and Jürgens, 1993). Analysis of these mutants revealed a tight correlation between the formation of short stretches of vascular tissue and the development of corresponding hypocotyl stumps (T.Berleth, unpublished data). Thus, vascular strand formation and basal organogenesis in the embryo might be directed by common underlying apical–basal cues.

The precise role of auxin in apical–basal signaling, suggested by a number of classical experiments, remains to be established. This developmental role may have not been fully addressed by genetic analyses of auxin functions, since auxin perception mutants have mainly been identified based on reduced auxin responses of adult plants (for review see Estelle and Klee, 1994; Hobbie et al., 1994). Experiments involving the local application of indole-3-acetic acid (IAA, the major form of auxin in higher plants) and chemical inhibition of auxin transport have implicated an apical–basal flux of auxin in the formation of vascular strands (reviewed in Sachs, 1981). Furthermore, seedling defects similar to those of mp mutants have recently been described for the Arabidopsis mutant auxin resistant 6 (Hobbie, 1997) and impaired embryonic symmetry, similar to cotyledon fusions observed in mp embryos, have been described for Brassica juncacea embryos treated with chemical inhibitors of auxin transport (Liu et al., 1993). Moreover, cotyledon fusions as well as [mp related (Przemeck et al., 1996)] spike-like inflorescences have been reported for the Arabidopsis mutant pin formed (Okada et al., 1991). The pin formed mutant is impaired in the polar transport of auxin, as is the mp mutant (Przemeck et al., 1996). These correlations, as well as the recently demonstrated capacity of the MP (IAA24) gene product to bind to functional auxin responsive promoter elements (Ulmasov et al., 1997), suggest that developmental auxin signals could be relayed by the MP gene product. The possibility can be tested once the authentic target genes of the MP product have been identified. Notably, a number of structurally related presumptive transcription factors have recently been identified (Ulmasov et al., 1997). The MP gene constitutes a member of this class of genes with genetically defined functions, and its further characterization should facilitate the analysis of the molecular signaling context (including the role of auxin), as well as genetic dissection of embryo and vascular development.

Materials and methods

High-resolution mapping of the MP region

Cosmid clones homologous to YAC CIC8C5 (Creusot et al., 1995) were obtained by screening a genomic library from Arabidopsis ecotype Landsberg erecta (Ler, Meyer et al., 1994) with DNA of this YAC eluted from pulsed-field gels. Nine cosmid clones were used to search for RFLPs between Arabidopsis ecotypes Landsberg erecta and Niederzenz. Six RFLPs were detected and mapped relative to meiotic recombination breakpoints in the region (data not shown). A local chromosome walk was initiated bidirectionally from a cosegregating cosmid (0/1796 recombinant meiotic products) aligning cosmid and BAC clones [from EcoRI partially digested Columbia-0 (Col-0) genomic DNA; a not-ordered small-insert library was kindly provided by T.Altmann, Golm].

Isolation of cDNAs

Positive MP cDNA clones were present at a frequency of 4×10−6 in a library from etiolated seedlings (Kieber et al., 1993). A likely full-length MP cDNA sequence was obtained by RACE PCR extension (5′3′-RACE-kit, Boehringer) of the longest available cDNA clone (KL1, 2.7 kbp). The length of the cDNA sequence (3.1 kbp) is consistent with the transcript size detected on poly(A)+ RNA blots (∼3.2 kbp).

Sequencing of mutant alleles

The extended cDNA and genomic DNA sequences of Ler, as well as Col-0 wild-type strains and of mutant alleles U252, CSH1, T370, G92 (Ler background), BS1354 and G33 (Col-0 background), were determined by direct sequencing of PCR products generated with Pfu DNA polymerase (Stratagene, LaJolla) from two independent DNA preparations. The phenotype and origin of some of the mutant lines have been described in Berleth and Jürgens (1993) and Przemeck et al. (1996).

Transient transformation of onion epidermis cells

Constructs for transient expression were generated by inserting MP cDNA fragments in-frame between a translational start ATG and the GUS reporter gene moiety in vector pNT160 (Boehm et al., 1995). Transformation of onion epidermis cells and histochemical staining was performed as described (Varagona et al., 1992) using a PDS1000 helium particle gun (Bio-Rad, Hercules).

In situ localization of transcripts in tissue sections

To synthesize the MP antisense probe, three cDNA fragments corresponding to nucleotide residues 210–825, 1281–1979 and 2659–3102 (maximum similarity to any data base sequence: 38% within a small interval) were subcloned in pSP72 (Promega), linearized by digestion with EcoRI and transcribed with SP6 polymerase in the presence of [35S]-UTP. Preparations of tissue sections, hybridization and exposure were performed as described (Drews et al., 1991). The sections were exposed with emulsion for ∼3 weeks.

Accession numbers

The accession numbers are AFO37228 and AFO37229.


We thank T.Altmann, J.Ecker, E.Grill, G.Jürgens, J.Kieber and E.M. Meyerowitz for molecular markers, technical advice, genomic and cDNA libraries, respectively; B.Scheres and R.Sung for mutant mp alleles and R.Kahmann for generous support at the institute. We are grateful to J.Müller for technical assistance, M.Strasser for help in developing a suitable in situ hybridization set-up and C.Gebhard for introducing us to the high-resolution Southern blotting technique. We would also like to thank E.Grill, R.Kahmann, R.Kunze, W.Lukowitz, S.Ploense, B.Scheres, D.Tautz and S.C.deVries for helpful suggestions on the manuscript. C.S.H. was supported by a pre-doctoral fellowship awarded by the University of Munich. This work was supported by grant Be1374 from the Deutsche Forschungsgemeinschaft (awarded to T.B.).