ETTIN patterns the Arabidopsis floral meristem and reproductive organs

In higher plants, proper floral development requires the coordinated activity of a number of genes that control the patterning of organ type, organ number and organ form (Weigel, 1995; Yanofsky, 1995). In Arabidopsis, these genes have been classified into those functioning early in the establishment of floral meristem identity such as LEAFY (LFY), APETELA1 (AP1), and APETELA2 (AP2), and those acting later during the patterning of organ identity such as APETELA3 (AP3), PISTILLATA (PI) and AGAMOUS (AG) (Weigel and Meyerowitz, 1994). Mutations in these genes affect the floral versus infloresence identity of floral shoots and/or the identity of the individual floral organs. Most of these genes are thought to encode transcription factors (Weigel, 1995; Yanofsky, 1995).

There are a number of other mutations which affect organ number, organ shape and regional differentiation within floral organs without changes in the identity of either the floral shoot or its component organs. For example, mutations in CLAVATA1 (CLV1), CLAVATA3 (CLV3), ETTIN (ETT), and PERIANTHIA (PAN) increase organ number within the flower, whereas mutations in TOUSLED (TSL) decrease organ number and organ size (Clark et al., 1993, 1995, and 1997; Roe et al, 1993 and 1997; Running and Meyerowitz, 1996; Sessions and Zambryski, 1995; Sessions, 1997). These genes likely act to pattern growth, cell division and regional differentiation during the acquisition of identity within the developing flower. CLV1 functions partially redundantly with meristem identity genes, whereas PAN and TSL function independently of the meristem and organ identity genes (Clark et al., 1993; Roe et al, 1997; Running and Meyerowitz, 1996).

Here we describe the isolation of the ETT gene and its expression pattern during early floral development. ETT encodes a protein that is predicted to be nuclear localized and is homologous to DNA binding proteins which bind to auxin response elements (AREs). ETT mRNA is detected in a complex pattern throughout early flower development and is consistent with the proposed role for ETT in patterning: (i) perianth organ number, and (ii) stamen and carpel form. Early ETT expression in the gynoecium suggests that the protein participates in the patterning of abaxial tissues and the establishment of apical and basal boundaries in the primordium. Interestingly, the transcriptional activation of ETT is independent of known meristem and organ identity functions. Lastly, genetic analyses indicate that ETT functions independently of CLV and redundantly with PA N to control floral organ number.

ett alleles and plant growth conditions were similar to those described by Sessions and Zambryski (1995). ag-5 was isolated and provided by Eva Huala (Roe et al., 1997). All other mutant alleles were obtained from the laboratory of Elliot Meyerowitz (California Institute of Technology).

Fixation, drying and viewing are the same as Sessions (1997).

Plasmids, strains and cloning details, as well as data not shown are available upon request. ETT was isolated using the ett-1 allele. ett-1 was identified in a large screen of T-DNA mutagenized seed (#7581; Feldmann, 1991; Azpiroz-Leehan and Feldmann, 1997) and shown by Southern analysis with T-DNA border-specific probes to contain a single T-DNA insert encoding kanamyacin resistance. The ett mutant phenotype was completely linked to the T-DNA and kanamyacin resistance (0 recombinants in 50 homozygous ett-1/ett-1 plants) in F₂ backcross populations. Left border plasmid rescue of the ett-1 T-DNA border generated a plant specific 0.4 kb StyI/EcoRI fragment that was used to screen a Wassilewskija (WsO) lambda genomic library (Roe et al., 1993). Five clones spanning 25 kb were isolated. Southern analysis mapped the 0.4 kb StyI/EcoRI border fragment to the center of the 12 kb clone ASL2. Clone ASL2 was inserted into transformation vector pSLJ6991 (Jones et al., 1992) as a SacI fragment and introduced directly into a suppressed ett-1 line (see below) by vacuum mediated transformation. ASL2 rescued the ett-1 phenotype and subsequently rescued ett2 and ett-3 in crosses, demonstrating that ETT resides within ASL2. Two overlapping fragments from ASL2, one spanning the predicted insertion site in ett-1, were used to screen a Landsberg erecta (LaO) floral and infloresence specific lambda ZAP cDNA library (Weigel et al., 1992). The longest (2.1 kb) of three similar clones, designated number 5, was purified as a pBluescript plasmid (pAS13), and sequenced in full (GenBank accession no. AF007788). cDNA 5 was used in northern analysis to show the presence of a 2.3 kb transcript in wild-type, ett-2, and ett-3 infloresence tips which was absent in ett-1. Sequencing primers derived from the cDNA 5 sequence were used to initiate sequencing of the complete 4.5 kb coding region of ETT genomic DNA from ASL2, corresponding to 300 nucleotides upstream of the predicted transcription start to 100 bp after the predicted translation stop. Alignment of the cDNA 5 sequence with ASL2 sequence suggests that the translation start site lies 77 bp downstream of the 5′ end of cDNA 5. The transcription start site is predicted to lie 341 bp upstream of the predicted translation start site based on RACE and primer extension analysis. The insertion sight of the left border of the T-DNA in ett-1 was determined by sequencing the left border rescue plasmid. 4.0 kb of ett-2, ett-3 and ett-4 covering the complete genomic region was amplified in two fragments by PCR (Hi Fidelity PCR Kit, Boehringer Mannheim), sequenced in full, and lesions found in fragments cloned from two independent PCR reactions listed in Fig. 2. ett-2 was recovered from T-DNA mutagenized WsO seed (#1537); ett-3 and ett-4 were isolated from ethylmethane sulfonate (EMS)-treated LaO seed and kindly provided by David Smyth and John Alvarez (Monash University).

In situ hybridization using digoxigenin-labeled probes and alkaline phosphatase detection was performed according to the method of Drews (1995) and according to the manufacturer’s (Boehringer Mannheim) directions. ett-1 plants were used as negative controls. 5′ and 3′ probes were transcribed from pBluescript subclones of pAS13 (1.6 kb EcoRI (5′), and 0.8 kb BamHI/XbaI (3′).

Double mutants were made using the null ett-1 allele which is in the WsO ecotype. With the exception of lfy-1 and ag-5 which are in the Columbia (ColO) ecotype, the remaining mutant alleles (ap1-1, ap22, pi-1, ap3-1, ag-1, clv1-1, clv3-1) are in LaO. Control outcrosses of ett-1 identified ecotype specific suppressors of ett-1 in ColO and LaO, which segregate as single recessive loci in F₂ outcross populations: 25% of F₂ett-1 plants have a suppressed gynoecium phenotype, allowing more valve formation and some seed set. The loci imparting suppression of ett-1 in LaO and ColO have not been mapped. In F₂ double mutant populations these suppressor loci segregated independently of ett-1 and all other mutations with the exception of lfy-1 and pi-1, which both reside on chromosome 5. Suppressed phenotypes in double mutant populations were distinct and additive, and were not scored in the phenotypic analyses presented.

ett-1 was crossed as a male onto the different homozygous mutants. Individuals doubly homozygous for ett-1 and the other mutation of interest were identified in five different ways: (i) as a new/additive phenotype segregating 1/16 in F₂populations (ap2-2 ett-1, ap3-1 ett-1, ett-1 pi-1, ett-1 pan-1); (ii) by Southern analysis of individual F₂ mutants to detect ett-1 specific T-DNA induced RFLPs (ett-1 lfy-1, ag-1 ett-1, ag-5 ett-1); (iii) in F₃ populations from F₂ suppressed ett-1 individuals that were heterozygous for the other mutation of interest (ett-1 lfy-1, ag-1 ett-1); (iv) in F₃ populations from F₂ mutants that were heterozygous for ett-1 (ap1-1 ett-1, clv11 ett-1, clv3-1 ett-1, ett-1 pan-1); and (v) by testcrosses (ett-1 pan1). The ap1-1 enhancing cauliflower-1 (cal-1) mutation resident in WsO (Bowman, 1993) was not present in the families segregating for ap1-1 ett-1 used here.

Wild-type Arabidopsis flowers are composed of four sepals, four petals, six stamens and a bicarpelate gynoecium (Fig. 1A). ett mutations only affect the flower, and are pleiotropic: sepal and petal number are increased (Fig. 1B), stamen number and anther form are decreased (Fig. 1D,E), and the proper differentiation of gynoecial tissues is significantly altered (Fig. 1GK; Sessions and Zambryski, 1995; Sessions, 1997). ett associated decreases in stamen number are correlated with aberrant or failed initation of medial stamen primordia (Sessions, 1997), whereas the anther defects include a failure in the formation of the interthecal furrow which is often partially formed or missing on the anthers of ett medial stamen (Fig. 1C-E). The anatomy of tissues and vascular bundles is normal in ett sepals, petals and stamens (not shown).

ett gynoecium phenotypes are allele-strength dependent and involve the aberrant development of tissues in place of the ovary. Phenotypes for each allele vary. Wild-type gynoecia consist of a stigma-and-style-capped bilocular ovary on an unelongated internode. The ovary is the largest region of the wild-type gynoecium and is composed abaxially of 2 valves laterally and 2 furrows medially. The phenotypes of ett-2, ett-3 and ett1 gynoecia form a continuum of decreasing ETT function in which valve tissue is removed basally (Fig. 1G-K). Lost valve is replaced basally by structures intermediate between abaxial style and internode, and medially (between the valves) by adaxial style tissue (Fig. 1F-K; Sessions and Zambryski, 1995; Sessions 1997). These phenotypes have been interpreted as resulting from the basalizing of a hypothetical distal abaxial-adaxial boundary, and the raising of a hypothetical proximal valve forming boundary on the gynoecium primordium, in an allele-strength dependent manner (Sessions, 1997). Unlike the sepals, petals and stamen, vascular patterning and anatomy are altered in ett gynoecia (Sessions and Zambryski, 1995).

An ETT genomic DNA clone was isolated using the T-DNA tagged ett-1 allele, and shown to complement ett-1, ett-2 and ett-3 (see Materials and Methods). The 2.3 kb ETT transcript has the potential to encode a 608 amino acid protein with a predicted molecular mass of 66.5 kDa (Fig. 2). The protein contains two serine rich regions and a putative bipartite nuclear localization signal (NLS; Robbins et al., 1991). Furthermore, the N-terminal half of ETT shows homology with the DNAbinding domains of the transcription factor ARF1 and the related protein IAA24 (Ulmasov et al., 1997; Kim et al. 1997). These two proteins have been implicated in mediating responses at auxin-regulated promoters. Moreover, using the ARF1 sequence, a BAC containing the ETT locus was identified in GenBank (U78721); a cDNA isolated with this BAC sequence was called ARF3 (Ulmasov et al., 1997). The sequence of ETT and ARF3 are identical with the exception of 5 nucleotides. In addition, an 84 amino acid region beginning at position 155 of the ETT protein and containing the NLS is highly similar to an EST from rice (D40316, 93% identity). The C-terminal half of ETT is unique and of unknown function.

Four ett alleles were sequenced and found to contain lesions consistent with the phenotypic severity of each allele (Fig. 2). The strong ett-1 allele has a T-DNA inserted into exon 2 and lacks the wild-type 2.3 kb transcript as determined by northern blot and in situ hybridization analyses (not shown). The weak ett-2 allele has a single bp change that leads to a conservative Arg to Lys substitution at amino acid 247, and also affects splicing in some ett-2 transcripts (J. N. and P. Z., unpublished). The improperly spliced transcript introduces the stop codon contained in intron 5 resulting in a truncated polypeptide. The intermediate strength ett-3 and ett-4 alleles contain nonsense mutations in exons 8 and 9, respectively, downstream of the putative DNA binding domain.

To determine when and where ETT is expressed during flower development, in situ hybridization was performed on floral tissues spanning the 13 developmental stages (Smyth et al., 1990). ETT RNA is first detected before stage 1 in the infloresence meristem (IM) in groups of cells which are developing as new FMs (floral meristems; Fig. 3A-C). The base of the expression domain tapers and joins the infloresence axis procambium (Fig. 3A-C). During stage 2, expression resolves to procambial tissues in the pedicel, and a domain in the future receptacle (Fig. 3A-C). This expression appears to mark the sites of vascular differentiation within the floral pedicel and receptacle (Fig. 3C). In addition, a second patch of expression is detected towards the apex of stage 2 FMs (‘ii’ in Fig. 3C) and this patch ultimately expands to form a thick ring in the terminal meristem (presumptive gynoecium) during stage 5 (Fig. 3D-F). During stage 3 and 4, expression in the pedicel remains restricted to the pedicel vasculature in cells that lie between differentiating phloem and xylem elements (not shown), and appears to be high in presumptive petal and stamen primordia (Fig. 3D,E). The expression in the infloresence procambium also resolves to cells that lie between differentiating phloem andxylem elements (Fig. 3L). These results demonstrate that ETT is expressed in a complex pattern throughout early floral meristem formation and floral organ initiation.

ETT RNA is not detected in sepal primordia (Figs. 3D-H). ETT transcript is detected throughout petal primordia during stages 4-6, and becomes restricted to procambial cells during stages 7 and 8, and ceases by stage 9 (Fig. 3F,K). During stage 5, stamen primordia arise between the receptacle and gynoecium rings of expression, and show abaxial expression of ETT transcript from inception until stage 7 (Fig. 3F,G,H). During stages 7-9, ETT transcript is detected in the stamen vasculature and in four bands of cells within each anther: two adaxial stripes near the interthecal furrow and two abaxial stripes between the locules and the vasculature (Fig. 3H,J,K). Stamen vascular expression ceases during stage 9, before morphological differentiation of cell types within the bundles is visible.

Similar to stamen primordia, the gynoecium shows abaxial expression of ETT at inception (Fig. 3F,G). The inner (adaxial) edge of the ring of ETT expression in the terminal meristem marks the abaxial-adaxial boundary of the gynoecium primordium before it emerges from the FM (Fig. 3F,G). Expression from stages 5-8 is strictly in the abaxial cells of the gynoecium primordium (Fig. 3F-I). This expression is refined during stage 9 to cells within the four differentiating vascular strands that lie between phloem and xylem elements (Fig. 3J,K; not shown). This vasculature expression persists until stage 12 (not shown).

In wild-type plants the IM initiates lateral shoot meristems which normally develop as FMs. In meristem identity mutants lateral shoots assume a mixture of floral and infloresence meristem characters. To determine if any of these genes functions as an upstream regulator of ETT expression, double mutant analyses and expression studies were performed.

LFY is thought to be one of the earliest acting genes in the floral meristem identity program because of loss-of-function phenotypes which convert lateral floral shoots toward an infloresence identity (Huala and Sussex, 1992; Weigel et al., 1992). Usually only the late initiated, most apical lateral shoots on strong lfy mutants resemble normal flowers, and even these lack petals and stamens (Fig. 4A; Huala and Sussex, 1992; Weigel and Meyerowitz, 1992). ett-1 lfy-1 plants differ from lfy-1 plants only in the terminal gynoecia of late initiated lateral shoots, which express the ett phenotype (Fig. 4B). Thus, lfy-1 is largely epistatic to ett-1 except in the most floral like shoots, suggesting that ETT is not active early in the development of lfy-1 infloresences and lateral shoots. ETT expression however, is largely normal in a lfy-1 mutant indicating that LFY is not required for the early transcriptional activation of ETT in the IM and in lateral meristems (Fig. 4C). Thus some aspect of the post-transcriptional function of ETT is dependent on LFY.

AP1 appears to function in concert with LFY in the control of FM identity (Bowman et al., 1993; Mandel and Yanofsky, 1995; Weigel and Nilsson, 1995). ap1 mutations cause the basal nodes of lateral shoots to have infloresence-like identity, leading to the formation of bracts and secondary floral shoots where sepals and petals would normally develop (Fig. 4D). ap1-1 ett-1 lateral shoots have w1 and w2 phenotypes that are identical to those of ap1-1, and w3 and w4 phenotypes which are identical to those of ett-1 (Fig. 4E). ap1-1 is epistatic to ett-1 in that no more than four w1 organs are initiated in ap1-1 ett-1 lateral shoots (not shown). Although it sometimes may appear that five w1 green bract-like organs are present in mature ap1-1 and ap1-1 ett-1 flowers (Figs. 4D, E), the extra organ develops from the positions of w2 petal primordia (Bowman et al., 1993). ETT thus appears to not be functioning in w1/w2 of ap1-1 lateral shoots. Similar to the case with lfy-1, ETT expression appears largely normal in ap1-1 IMs and FMs (Fig. 4F), suggesting that AP1 is required at the posttranscriptional level for ETT activity.

Since LFY, AP1 and CAL, have been shown to be partially redundant coregulators of meristem identity, expression of ETT was assayed in ap1-1 lfy-1 and ap1-1 cal-1 double mutants. ETT expression in lateral meristems occurs in both double mutants, supporting the conclusion that ETT transcriptional activation occurs independently of meristem identity genes (Fig. 4M; not shown).

AP2 is defined as a meristem identity gene due to secondary flower production in ap2-1 flowers, and the enhancement of lfy and ap1 infloresence characteristics by ap2 alleles (Irish and Sussex, 1990; Huala and Sussex, 1992; Bowman et al., 1993; Okamuro et al., 1997). Strong mutations in AP2 such as ap22, result in the development of medial w1 carpels in place of sepals, the absence of w2 petals, loss of w3 stamens, and occasionally unfused w4 carpels (Fig. 4G; Bowman et al., 1991). ap2-2 is largely epistatic to ett-1 in w1-w3 of ap2-2 ett-1 flowers (Fig. 4H). ett-1 does not affect organ number in w1 of ap2-2 flowers, but does affect development of w1 carpel margins (Fig. 4I-J). ap2-2 w1 medial carpels generally have central valve tissue bounded laterally by a marginal flap of medial ovary and stigmatic tissue, and a transmitting tract-covered submarginal placenta (Fig. 4I). ap2-2 ett-1 w1 medial carpels lack the marginal flaps, and valve tissue abuts the transmitting tract-covered submarginal placenta (Fig. 4J).

ap2-2 ett-1 double mutants suggest that although ETT is not active in determining organ number in ap2-2, it is required in patterning differentiation within the margins of w1 ap2-2 carpels (Table 1; Fig. 4J). The normal valve development in ap2-2 ett-1 w1 medial carpels suggests ETT is not needed for the development of valve tissue per se, and that its role in normal w4 gynoecium development is probably to position where valve cell types will form. AP2 does not appear to be involved in the early transcriptional activation of ETT, since expression is normal in ap2-2 meristems (Fig. 4K). While ETT is not normally expressed in w1 sepals, in ap2-2 ETT RNA appears in the abaxial layers of w1 carpel primordia, suggesting that AP2 is somehow involved in repression of ETT expression in w1 (Fig. 4L). It is unclear why ETT is transcribed in the abaxial layers of the valve regions of w1 ap2-2 carpels when it only seems to be functioning in the margins.

Organ identity genes have been typed into a, b and c classes (Bowman et al., 1991). We examined the potential interaction of ETT with members of each class genetically and by in situ hybridization.

AP1 and AP2, in addition to being classified as meristem identity genes, are also considered ‘a’ class organ identity genes. As described above, the organ identity functions of AP1 and AP2 appear to act independently of ETT, since ett-1 does not affect organ identity in ap1 and ap2 mutants (Fig. 4E,H; Table 1).

‘b’ class genes are represented by AP3 and PI, mutations in which lead to the replacement of petals by sepals and stamens by carpels. Flowers from plants homozygous for the weak ap3-1 allele have sepals in w2 and carpeloid stamens in w3 (Fig. 5A,B; Bowman et al., 1989). Carpeloid stamens in ap3-1 flowers have a variable phenotype but are generally mosaics in which individual anther locules develop with carpeloid features, most noticeably placentae (Bowman et al., 1989; Fig. 5B). ap3-1 ett-1 flowers have a largely additive phenotype of 5 to 6 sepals in each of w1 and w2, w3 organs showing reduced locule formation and generally lacking all placental features, and a w4 ett gynoecium (Table 1; Fig. 5C,D). Most notable of ap3-1 ett-1 flowers is the development in w3 of anther loculelacking club shaped organs covered in anther-like tissue (Table 1; Fig. 5C,D). This phenotype can also be interpreted as additive since both ap3-1 and ett-1 diminish locule formation. Flowers from plants homozygous for the strong pi-1 allele have sepals in w1 and w2 and carpeloid organs in w3 and w4 (Fig. 5E,F; Hill and Lord, 1989). pi-1 w3 organs can be either free from or fused to the central w4 carpels (Fig. 5E,F; Hill and Lord, 1989). ett-1 pi-1 flowers have an additive phenotype of 5-6 sepals in each of w1 and w2 and ett-like carpels which lack valve tissues in w3 and w4 (Fig. 5G,H). The additive phenotypes of ett-1-‘b’ class double mutants suggest that ETT functions independently of ‘b’ functions during normal petal and stamen development. Additionally, we have found that ett-1 shows additive interactions with mutations in genes whose wild-type products regulate ‘b’ function activities, including unusual floral organs (ufo)-2, and superman (sup)-1, further suggesting independence of ETT and ‘b’ function activity (not shown; Bowman et al., 1992; Levin and Meyerowitz, 1995).

AG is the only identified ‘c’ class gene in Arabidopsis. ag mutations cause the development of petals and sepals in place of w3 stamens and w4 carpels (Fig. 5I; Bowman et al., 1989).

Additionally ag flowers are indeterminate and continue nitiating organs inside the w4 sepals in the whorl-type pattern (petals, petals, sepals)_n (Fig. 5I). Strong mutations in AG, such as ag-1, are epistatic to ett-1 in w3 and w4 (Table 1; Fig. 5J). Weaker alleles of ag, such as ag-5, lead to less severe organ identity transformations (Roe et al., 1997) and are enhanced by ett-1 (Fig. 5K, L). Whereas ag-1 ett-1 flowers suggest that ETT is not active in ag-1, ag-5 ett-1 flowers indicate that ETT is necessary for full AG activity. In situ hybridization suggests that AG plays no role in the transcriptional regulation of ETT (Fig. 5M-O). Surprisingly, ETT shows normal abaxial expression in ag-1 w3 and w4 organs from inception, as well as abaxial expression in subsequently initiated organs (Fig. 5M-O). The abaxial expression of ETT in ag-1 w3-w6 primordia supports the view that ag-1 affects organ identity after initiation of normal primordia (Crone and Lord, 1994). In particular, ag-1 w4 primordia which will develop as sepals, express ETT abaxially, similar to carpel primordia, and not sepal primordia, which normally lack ETT expression at inception.

CLV1 and CLV3 are proposed to function together to promote differentiation of cells in shoot meristems (Clark et al., 1993, 1995, 1997). Mutations in either gene lead to increases in meristem size, and to increases in organ number in all floral whorls (Fig. 6A; Table 1). clv1-1 and clv3-1 each act additively in double mutant combination with ett-1 (Fig. 6B; Table 1; not shown for clv1-1 ett-1). clv3-1 ett-1 flowers had slight increases in organ number in each whorl compared to single mutants (Table 1). Stamens and carpels in both double mutants show ett-1like alterations in regional differentiation of cell types. Notably, clv3-1 ett-1 flowers show a synergistic increase in the number of w3 staminodes (reduced stamen), suggesting an interactive role for CLV and ETT in the promotion of stamen development (Table 1; Fig. 6B). The indeterminacy caused by clv mutations (Clark et al., 1993,1995) which is normally contained within the w4 gynoecium is exposed in clv3-1 ett-1 flowers due to the splitting of the style and stigma caused by ett-1 (Fig. 6B).

PAN is thought to function to promote bilateral symmetry within the floral meristem (Running and Meyerowitz, 1996). pan mutations change FM symmetry from bilateral to radial, which, like ett mutations, increase sepal and petal number, and decreases stamen number (Fig. 6C). ett and pan mutants differ in gynoecium defects: ett mutants have tissue patterning defects within-carpels and pan mutants have increases in carpel number (Running and Meyerowitz, 1996). pan-1 mutant FMs develop similarly to ett-1 mutants in the increase in adaxial sepal and petal number and loss of stamen primordia (Fig. 6E,F,H,I,K,L). pan gynoecium development is similar to wild-type with the exception of early trumpet-like as opposed to cylindrical growth of the gynoecium primordium (Running and Meyerowitz, 1996).

ett-1 pan-1 double mutants show synergistic phenotypes, the most dramatic of which is a loss in the proper spacing and an increase in the number of petal primordia within w2 (Fig. 6D; Table 1). Flower development in ett-1 pan −1 double mutants diverges from that of ett-1 and pan-1 single mutants at stage 3 when the sepal primordia initiate (Fig. 6G). Fewer sepal primordia initiate in ett-1 pan-1 FMs. These sepals are variably clustered on one side of the FM (Fig. 6G). Aberrant sepal initiation is followed by dramatic proliferation of the remaining FM (Fig. 6G). This proliferation is followed by the initiation of multiple petal primordia, a diminished number of stamen primordia and a trumpet-shaped gynoecium primordium (Fig. 6J, M). Development of sepals, stamens and the gynoecium is also synergistically impaired in ett-1 pan-1 double mutants: sepal and stamen primordia often grow into narrow reduced organs, and the gynoecium usually lacks ovules and appears like a reduced ett-1 gynoecium (not shown).

The enhancement of ett-1 and pan phenotypes in double mutant combination, and the similar w1-w3 phenotypes of individual ett and pan mutants suggest that ETT and PAN function redundantly in the within-a-whorl spacing of sepal and petal primordia, the initiation of stamen primordia, as well as placental development within the gynoecium. ETT transcription in pan FMs and IMs is normal (not shown).

Based on ett mutant phenotypes and the expression pattern of ETT transcripts we propose that ETT has a dynamic role in patterning development in groups of cells within floral meristems and reproductive organs. In early patterning, ETT functions in determining the number of organ primordia, whereas later it is involved in the outgrowth of and patterning of tissues within organ primordia. Mechanistically, how the ETT gene product achieves this is unclear, though it is likely to function as a transcription factor. The homology of ETT with the ARE-binding proteins ARF1 and IAA24 suggests that ETT may mediate auxin responses at the promoters of auxin regulated genes (Ulmasov et al., 1997; Kim et al., 1997).

The pleiotropic nature of ett mutants could result either from a requirement for ETT at an early stage of development which secondarily affects later stages, or from multiple requirements for ETT functioning throughout flower development. Expression patterns and the dissection of individual whorl functions in double mutant combination indicate that ETT functions at multiple times during flower development and that its activity within each whorl occurs independently of its functioning in other whorls. For example w1/w2 and w3/w4 ETT functions can be separated (ap1-1 ett-1, ag-1 ett-1), and w1/w2/w3 and w4 ETT functions can be separated (ap2-2 ett-1).

The early patterning role of ETT affects the number of sepal and petal primordia. ETT expression in the IM occurs before morphological appearance of the FM and resolves to a pattern by stage 3 which marks the presumptive sites of vascular development and future organ initiation. Expression in the future pedicel and receptacle regions of the FM appears last in cells between differentiating phloem and xylem elements. Defects in the anatomy of ett-1 pedicel vascular bundles, however, have not been detected.

The early reticulate expression somehow affects the positioning of the sites of sepal and petal initiation within the FM, and the sites of their vascular bundles within the pedicel and the receptacle, without affecting the size of the FM, or the differentiation of cell types within vascular bundles, sepals or petals. It is curious that sepal primordia do not express ETT transcript whereas petal primordia do, since both sepal and petal number are increased in ett mutants. This suggests that ETT functions within the stage 2 FM before sepal and petal primordia emergence to repress organ formation, perhaps by a mechanism related to lateral inhibition. ETT appears to perform this function redundantly with PAN (see below), but independently of the action of CLV genes, which act more globally in shoot meristems to promote the differentiation of cells. ETT also appears to control primordia number independently of the TSL protein kinase (Roe et al., 1997).

The later role of ETT affects the initiation of and patterning of tissues within stamen and carpel primordia. Stamen and carpel primordia express ETT abaxially from inception, and later within the developing vascular bundles. Early expression of ETT in w3 must function in part to organize individual primordia, since these primordia often fail to initiate in ett mutants (Sessions, 1997). This early expression appears to perform a function that requires CLV1 and CLV3, and that can be compensated for by PAN, to promote the outgrowth of normal primordia. CLV1 encodes a putative receptor kinase which is absent from stamen primordia but is expressed in the center of stage 4 FMs, perhaps overlapping with ETT expression (Clark et al., 1997).

Stage 7 stamen primordia additionally express ETT in four subepidermal vertical streaks in the anther. This anther expression apparently is essential to position where locule outgrowth and the interthecal groove will form as these functions are lost in ett-1. The stamen procambial cells express ETT transcript until stage 9, before visible differentiation of vascular cell types, although similar to the petals, stamen vasculature appears unaffected in ett mutants. That petal and stamen vasculature is normal in ett mutants suggests that ETT is not functioning in differentiation in these places, or that its function is covered by redundant factors in ett mutants.

The ett allelic series suggests that ETT patterns the gynoecium primordium in a dose-dependent manner. Abaxial expression of ETT in the gynoecium primordium from inception until stage 8 supports its predicted role in prepatterning the proper differentiation of tissues within the developing organ (Sessions and Zambryski, 1995). The expression data suggest that ETT acts in the abaxial walls of the primordium to perform three essential functions: (i) to promote formation of valve and ovary cell types, (ii) to repress formation of stylar and internode cell types and (iii) to pattern the sites of vascular differentiation and anatomy. One way ETT accomplishes this is by restricting the activity of the TSL protein kinase to the distal gynoecium primordium (Roe et al., 1997).

We have proposed a model in which the proper differentiation of tissues within the developing gynoecium occurs from two ringed boundaries established in the stage 6 gynoecium primordium (Sessions, 1997). Based on this model, ett mutations cause an apical (abaxial-adaxial) boundary to be lowered and a basal (valve forming) boundary to be raised on the stage 5/6 primordium (Sessions, 1997). This model is in part supported by ETT expression in the gynoecium primordium. For example, the inner edge of the ring of ETT-expressing cells in the stage 5 gynoecium primordium appears to mark the abaxial-adaxial boundary and the hypothetical apical boundary, arguing that ETT acts early to directly position this boundary. Similarly, the lower edge of ETT expression on the stage 5 gynoecium primordium appears to mark the proposed basal boundary. Thus, ETT could act to position these hypothetical developmental boundaries at the edges of it expression domain. Alternatively ETT could act throughout the gynoecium primordium (i.e. not entirely at the edges) to provide positional information.

ett-1 pi-1 flowers suggest that ETT functions similarly in w3 carpel primordia. However, the two boundary model is at odds with the phenotype of ap2-2 ett-1 w1 carpels, since ETT appears to be functioning only in the margins of these organs as opposed to throughout the primordium. Perhaps redundant factors are present in ap2-2 w1 carpel primordia which can promote the formation of valve tissue in the absence of ETT, or there is a different developmental basis for carpel development in w1.

Expression studies demonstrate that ETT is transcribed independently of meristem and organ identity genes. Recent studies indicate that LFY and AP1 are necessary and sufficient for flower formation within shoot primordia (Mandel and Yanofsky, 1995; Weigel and Nilsson, 1995), and that AG is necessary and sufficient for the formation of carpels (Mizukami and Ma, 1992), yet ETT transcriptional activation occurs in the absence of each of these functions. Additionally, loss of pairs of partially redundant meristem identity functions in ap1-1 lfy-1 and ap11 cal-1 plants does not appear to alter the early activation of ETT transcription in incipient and developing lateral meristems. This argues that other unidentified factors besides the known meristem and organ identity genes are involved in the transcriptional regulation governing flower development.

While expression studies demonstrate that ETT is expressed in the meristem and organ identity mutants lfy-1, ap1-1, ap22 and ag-1, double mutants suggest that ETT is not active in all whorls in these mutant backgrounds. Thus, meristem and organ identity genes although unnecessary for ETT transcriptional activation, are indeed necessary for ETT function. Additionally, ETT appears to be necessary for full AG function, since the partial function of the ag-5 allele requires ETT activity.

Several results suggest that ETT and PAN function redundantly to control radial patterning within w1-w3. First, the w1-w3 phenotype of both single mutants, including the early stage 14 floral ontogeny, is very similar. The developmental basis for stamen loss during stage 5 differs between the two mutants in that ett fails to initiate one of the four medial stamens while pan mutants initiate 5 equally spaced stamen primordia (Running and Meyerowitz, 1996; Sessions, 1997). Second, ett and pan mutations have similar genetic interactions in CLV and meristem and organ identity genes (this study; Running and Meyerowitz, 1996). Third, ett-1 pan-1 double mutants show different synergistic phenotypes in each whorl. This is best exemplified in w2 of ett-1 pan-1 flowers in the formation of petal primordia in all available positions in the whorl. ETT and PAN seem to function independently of each other, since each is active in a mutant of the other. Since ett and pan single mutants do show clear differences, the extent of the redundancy is unclear. Future experiments expressing ETT in a pan background under constitutive and spatially refined promoters should help to clarify this relationship.

ETT provides a critical function in patterning groups of cells within the FM and reproductive organs. Understanding the nature of this patterning function will be aided in the future by gain of function ETT alleles, and experiments which establish whether ETT acts as a transcription factor and/or mediates auxin-based signals. Transcriptional activation of ETT occurs independently of meristem and organ identity genes, suggesting that other unidentified factors are necessary for flower development. ETT expression also implies that primordium initiation and vascular patterning are coincident events, and that the differentiation of tissues in mature organs is partially patterned in the meristem before primordia become morphologically distinct from the meristem.

We thank John Alvarez and David Smyth for ett-3 and ett-4, Eva Huala for ag-5, Elliot Meyerowitz for clv1-1, clv3-1, and other single meristem and organ identity mutants, Ove Nilsson and Detlef Weigel for ap1-1 (lfy-6/+), James Keddie for pSLJ6991, the Berkeley Electron Microscope Lab, Steve Ruzin and the NSF center for Plant Developmental Biology, and Tim Durfee and Fred Hempel for discussions and critical reading of the manuscript. This work was supported by Department of Energy grant 88ER13882 to P. C. Z.