Development 104, 195-203 (1988)Printed in Great Britain © The Company of Biologists Limited 1988

195

Isolation and characterization of novel mutants of Arabidopsis thaliana

defective in flower development

MASAKO K. KOMAKI1, KIYOTAKA OKADA1, EISHO N1SHINO2 and YOSHIRO SHIMURA1

1 Division of Cellular Communication, National Institute for Basic Biology, Okazaki 444, Japan2Biological Laboratory, College of Arts and Sciences, Chiba University, Chiba 260, Japan

Summary

We have isolated a number of mutants of Arabidopsisthaliana, a member of the mustard family, that havedefects in flower development and morphogenesis. Ofthese, five mutants have been extensively character-ized. Two mutants (Fl-40, Fl-48) lacking petals showhomeotic conversion of sepals to carpels. One mutant(Fl-54) displays highly variable phenotypes, includingseveral types of homeotic variations, loss or distortedpositions of the floral organs as well as abnormalstructures on the inflorescence. Two other mutants(Fl-82, Fl-89) show aberrant structures in the pistils.

Genetic analyses have revealed that these mutationsare single and recessive, except for one mutant whosemutational loci still remain to be determined. Thesemutants may prove useful for the analysis of thegenetic control of flower development and morphogen-esis in the higher plant.

Key words: Arabidopsis thaliana, flower development andmorphogenesis, flower mutants, homeotic mutation,deletion of floral organs, genetic analysis.

Introduction

The flower is a reproductive organ in angiosperms,consisting of gynoecium (pistil), androecium (sta-men) and other floral organs (petal, sepal). Thesefloral organs are arranged in whorls or concentriccircles. Thus far, the developmental and morphogen-etic processes of flowers have been studied mainlyusing organographic and embryological approachesand very little study has been undertaken genetically.The difficulties encountered in the genetic approachfor studying development and morphogenesis inhigher plants are primarily due to their long lifecycles, the large genomes and the requirement forlarge growing facilities such as farms and green-houses.

Arabidopsis thaliana (L.) Heynh. is a small cru-cifer, which has various properties that make it aconvenient plant for molecular genetic research (forreview see R6dei, 1975; Meyerowitz & Pruitt, 1985;Estelle & Somerville, 1986; Meyerowitz, 1987). It hasa generation time of 6-8 weeks and a genome size(7xlO7 base pairs per haploid) of about a hundredthof the size of most higher plants. The plant is small(20-30 cm in height) and easy to grow in laboratories.

The procedures for mutagenesis, Ti plasmid-me-diated transformation and plant regeneration arebeing studied. Using this flowering plant, we haveattempted to take a genetic approach to understandthe process of flower development and morphogen-esis. The very first and most fundamental step in thestudy is to isolate and characterize mutants that havedefects in flower development and morphogenesis. Itis expected that analyses of such mutants will revealthe genes responsible for those mutations and willeventually lead to an understanding of the molecularmechanisms involved in the developmental pro-cesses.

The flower morphology of Arabidopsis has beenextensively studied (Miiller, 1961) and shares a com-mon structure with that of Brassicaceae (Cruciferae),the mustard family. Several mutants of Arabidopsishaving an altered flower morphology have alreadybeen reported (McKelvie, 1962; Koornneef et al. 1980,1983; Pruitt et al. 1987; Haughn & Somerville, 1988).Here we report the isolation of novel mutants defec-tive in flower development and morphogenesis, andalso the results of detailed examinations of the floralstructures and genetic analyses of several mutants.

196 M. K. Komaki and others

Materials and methods

Plant linesAn Arabidopsis thaliana wild-type strain, Landsberg(erecta), and mutant lines, M7 (apetala-1, clavata-1) andM10 (apetala-2, eceriferum-2), were obtained from Arabi-dopsis Information Service (Dr A. R. Kranz, BotanischesInstitute, J. W. Goethe-Universitat, Frankfurt am Main,FRG). Another mutant line, pistillate, was a gift from Dr E.M. Meyerowitz (Caltech). These mutations have beenmapped (Koornneef, 1987).

Cultivation of plantsPlants were grown according to the procedures of Dr C. R.Somerville (personal communication) under continuousillumination (24L) at 22 °C under the standard growthconditions. Some plants were grown under short-daygrowth conditions at 22 °C, where 8h illumination and 16 hdarkness (8L: 16D) alternate. Under the short-day growthconditions, plants become bigger than those grown underthe standard conditions, but flowering delays by about3-5 weeks.

Mutagenesis and mutant screeningThe wild-type seeds were mutagenized by soaking them in0-3% EMS (ethyl-methane-sulphonate) solution for16-24h according to the procedure of Dr C. R. Somerville(personal communication). The mutagenized seeds (Mlseeds) were sown, and the self-fertilized seeds (M2 seeds)were obtained. Mutants exhibiting abnormal floral struc-tures were isolated from the M2 plants.

Phenotype analysesTo characterize floral structure, more than 20 flowers werecarefully examined under a binocular microscope at lowmagnification. In order to examine morphological vari-ations, the flowers were selected from the lower and upperparts of more than five inflorescences of different plants.

Thin sections of flowers were prepared as follows:samples of the flowers were fixed by dipping overnight inBouin's fixative (75:25:5 (v) mixture of saturated picricacid, formalin and acetic acid), solidified in paraffin aftergradual dehydration, thin sectioned (5-10//m thickness)and stained with 0-5% (w/v) Azure-B solution in 0-1%sodium acetate buffer, pH4-6.

Genetic analysesMutant plants were crossed with each other or with wildtype by pollinating young unfertilized pistils by hand. WhenFi progeny generated from a cross between a mutant(homozygote) and a wild type showed the normal (wild)phenotype, the mutation was determined to be recessive.When the mutant was used as the male parent (pollen) in across, and when the mutation was inherited to the progeny,the mutation was judged to reside in the nuclear genome.When the mutation followed the laws of Mendelian segre-gation, namely plants with the recessive mutant phenotypeand plants with the normal phenotype were segregated in aratio of 1 to 3 in the F2 plants generated from self-fertilization of the F! plants, it was concluded that themutation was single.

Median plane

Inflorescence axis..

Pistil

Adaxial sepal

Petal

Outer stamen s \ \ \Q-0 .

Inner stamen—- '

Transverse plane

Lateral sepal (inner)

Abaxial sepal (outer)

Fig. 1. Diagram of Arabidopsis thaliana wild-type flower.

Results

Anatomy of the flower of wild type Arabidopsisthaliana

The flowers of Arabidopsis thaliana, like those ofother members of the mustard family, consist of foursepals surrounding and alternating with four whitepetals. Within the whorl of petals are six stamens(two short outer stamens and four long inner sta-mens) and a pistil having two carpels. Thus, thesefloral organs are arranged in concentric circles asillustrated diagrammatically in Fig. 1. It is worthnoting in this connection that although there has beenan alternative view that the pistils of Brassicaceaeoriginate from four carpels (Lawrence, 1951), wefollow the two-carpel hypothesis in this report forcharacterization of the pistil structure.

The floral organs are unambiguously distinguish-able from each other on the basis of their shape,colour and size. In addition, some specific surfacestructures could be used as the markers to identify theorgans. For example, the trichomes are found on theouter surface of the sepals, but not on other floralorgans. The branched trichomes are seen on bothsides of the leaves. Each floral organ also showsspecific inner structures, such as the patterns ofvascular bundles, specific cells and tissues. The struc-tural patterns of the flowers including the numbersand shapes of the floral organs are generally stableand not affected by growth conditions with theexception of the number of stamens. We have oc-casionally observed flowers with 4 or 5 stamens (13out of 44 flowers examined). The floral structures ofwild-type Arabidopsis thaliana described above havebeen used as the standards for screening and charac-terization of mutants. Using the mutagenesis pro-cedure described in Materials and Methods, we haveobtained a number of mutants among the M2 progeny

Flower mutants of Arabidopsis 197

Table 1. Average number of floral organs in flowers of the mutant strains of Arabidopsis thaliana

Strains

wild type

Fl-40

Fl-48

ap-2

Fl-54

Fl-82

H-89

Illuminationcondition*

24L

8L:16D

24L

8L: 16D

24L

8L: 16D

24L

8L: 16D

24L

8L: 16D

24L

8L: 16D

24L

8L: 16D

Number offlowers

examined

22

22

22

22

26

21

20

22

21

21

20

20

20

—

Pistil(carpels/pistil)

1 0(2-0)1-0

(2-0)1-0

(2-1)1-4*

(2-1)10

(2-1)l i t

(2-0)1-0

(2-0)M t

(2-0)10

(2-0)1-0

(2-0)1-0

(3-6)1-1

(3-3)1 0

(2-0)_

Stamen

5-7

5-5

1-6

3-7

0-8

3 0

5-3

4-8

4-7

5-0

6-2

6-4

5-4

—

Petal

4-0

4-0

0

0

0

0

3-8

3-6

1-8

2-4

4-2

4 1

3-4

_

Sepal(curved sepalt)

4 0(0)4-0(0)3-3

(2-2)4 0

(1-7)3-1

(2-1)3-5

(2-8)3-9

(0)3-7(0)3-8(0)3-3(0)4-2

(0)4-1(0)4-0(0)—

Remarks

normal structure

homeotic conversion of sepalsto carpels, loss of petals

homeotic conversion of sepalsto carpels, loss of petals

homeotic conversions of petalsto stamens and of sepals to leaves

various homeotic conversions,abnormal inflorescence

large pistils composed ofmany carpels

pistils with two stigmas

*24L, grown under continuous lighting. 8L: 16D, grown under 8h lighting and 16 h dark,t Curved sepal is the converted, ovule-bearing sepal. See text.t Axillary buds and pistils were counted.

that have defects in floral structures. Among themutants, five were chosen and extensively investi-gated. The characteristic morphological features andthe results of the genetic analyses of these mutantsare summarized in Tables 1 and 2, respectively.Detailed descriptions of the mutants are given sub-sequently.

Flower structures of the homeotic mutants, Fl-40 andFl-48The flowers of a mutant strain, Fl-40, have unusualstructures quite different from those of the wild type(Fig. 2A-C). There are substantial phenotypic vari-ations among the flowers even in one plant. Thetypical phenotypes of this mutant flower are illus-trated in Fig. 3A-F. Petals are completely absent andthere are fewer stamens and sepals (Table 1). In someflowers, neither the normal stamens nor the normalsepal can be found (Fig. 3A), whereas some otherflowers contain one or two stamens (Figs 2B, 3B,C).There are some flowers that have a pair of seeminglynormal sepals in the transverse position (Fig. 3C). Itappears that only the pistil maintains its normal

structure in this mutant, although the carpels are notfused in some flowers (Figs 2C, 3D,E).

The most remarkable feature of this mutant is thatall the flowers have two curved, thick, green struc-tures on the outermost circle at the median positionwith respect to the inflorescence axis. At the top ofthese structures, there is a dense growth of whiteunicellular hairs. The hairs look similar to the stigma-tic papillae of the normal pistils. On the inner surfaceof the structures, small light-green particles are at-tached with a short stalk along the margins of thestructures (Fig. 2B). The size (100-150 pan in diam-eter and 50-90 j/m in width) and the shape of theparticles are comparable to those of the ovules in theovary. Microscopic analyses of the thin sections(Fig. 2G,H) of the particles show that the pattern ofcells in the particles is identical to that of the ovules inthe ovary of the wild type, namely multiple cell layerssurround an embryo sac, which is located in thecentre, and the micropyle, which is at one end of theparticle (Fig. 21). The particles are attached to thebranches of the marginal vascular bundles that runalong the margins of the structures.

198 M. K. Komaki and others

These morphological analyses of this mutant showthat the novel structures carrying the ovule-likeparticles are structurally equivalent to, or possibly thesame as, carpels. The pattern of the vascular bundlesand that of particle attachment are consistent withthose of the theoretical primitive carpels proposed byEames (1931) (Fig. 3G). These observations led us toconclude that the outer pair of sepals is convertedhomeotically into a pair of carpels. The convertedsepals, however, differ from the normal carpels in thefollowing two aspects; (1) the mutant sepals havetrichomes on the outer surface like the normal sepals(Fig. 2B), whereas the normal carpels never beartrichomes, (2) the anther-like structures containingthe pollen are often attached to the side edge of theconverted sepals (Fig. 2B).

All the structural features of Fl-40 described aboveare also observed in another independently isolatedmutant, Fl-48 (Fig. 2D). However, the variantflowers without stamens appear more frequently inFl-48 than in the former mutant (Table 1). In thisconnection, the fertilization frequency of Fl-48 islower than that of Fl-40.

Genetic analyses of the Fl-40 and Fl-48 mutantsThe homozygous lines of Fl-40 and Fl-48 are eachexamined genetically by making crosses with severalstrains. The results are summarized in Table 2. Sinceeach of the Fl-40 and Fl-48 mutations followed thelaws of Mendelian segregation in the F2 generation,they represent single mutations on the nucleargenome and are recessive with regard to the wild-typeallele. When Fl-40 and Fl-48 were crossed with eachother, the flowers of the Fx progeny were indis-tinguishable from those of both parental strains,indicating that the mutations of Fl-40 and Fl-48 areallelic and codominant.

The two mutants were crossed further with severalmutant lines known to have abnormal floral struc-tures. The Fl-40 and Fl-48 mutants are non-allelicwith apetala-1 (ap-1), clavata-1 (clv-1) and pistillata(pi). However, the mutations of Fl-40 and Fl-48 areallelic with the apetala-2 (ap-2) mutation, which isknown to cause homeotic conversions of petals intostamens and sepals into leaves or bracts as describedbelow. The heterozygous plants from the reciprocalcrosses between the Fl-40 and ap-2 strains and be-tween the Fl-48 and ap-2 strains .always show thephenotypes of the typical ap-2 flowers. In the F2

generation, the hidden phenotypes of Fl-40 or Fl-48were segregated in a ratio of 3:1. Thus, the ap-2mutation is dominant over the Fl-40 and Fl-48 mu-tations.

Although the mutational loci are allelic, the floralstructures of Fl-40 and Fl-48 mutants are considerablydifferent from those of the ap-2 mutant (Figs 2E,

3H). The ap-2 mutant was originally reported byKoornneef (1980) as a mutant having reduced petalsand large sepals. Recently, the floral structure of thismutant has been studied extensively by Meyerowitzand colleagues (Pruitt et al. 1987, Meyerowitz, 1987)and by Haughn & Somerville (1988). They reportpartial or complete homeotic conversion of petals tostamens and of sepals to leaf-like structures. We haveconfirmed their observations. Unlike Fl-40 and Fl-48,the ap-2 mutant always carries four sepals, fourpetals, and four to six stamens (Table 1). The petalsof ap-2 show intermediate structures between thepetals and stamens of the wild-type strain. The petalsare small and wrinkled and carry yellow swellings likepollen sacs along their margins (Fig. 2F). The pollengrains as well as the pollen mother cells are found

Fig. 2. Flowers of Arabidopsis thaliana mutants. Plantswere grown under continuous illumination (24L), orunder short-day conditions (8L:16D, 8h light: 16h dark).(A) Wild type flower. 24L. x20. Bar, 800 fan.(B) Fl-40 mutant flower. 24L. an, anther; ov, ovules; st,stigmatic papillae; tr, trichomes. x22.(C) Fl-40 mutant flower. 24L. Arrows Indicate opencarpels that correspond to a pistil. X20.(D) Fl-48 mutant flower. 24L. an, anther; ov, ovules; st,stigmatic papillae. This flower lacks normal stamens.X27.(E) Apetala-2 mutant flower. 24L. an, anther-likestructure, tr, trichomes on inner and outer surfaces ofsepals. x22.(F) Apetala-2 mutant petals. A petal and a stamen of wildtype are also shown for comparison. Samples were fixedin Bouin's fixative dehydrated in butanol and examinedunder a dark-field microscope. 1, a wild type petal; 2,3,ap-2 mutant petals; 4, a wild-type stamen. Arrowsindicate immature pollen grains.(G) Transverse section of a wild-type flower, an, anther;ov, ovule; p, placenta; s, septum. Bar, 100 fan.(H) Transverse section of the converted sepal of the Fl̂ lOmutant, ov, ovules attached to the inner margins of theconverted sepal; vb, central sepal bundle; db, marginalsepal bundle; st, stigmatic papillae. Bar, 100 fan.(I) Longitudinal section of an ovule attached to theconverted sepal of the Fl-40 mutant, em, embryo sac; mp,micropyle; sk, stalk of ovule. Bar is 50/an.(J) Fl-40 mutant flower. 8L: 16D. Arrows indicateadventitious flowers. xl8.(K) Apetala-2 mutant flower. 8L: 16D. Arrow shows asmall extra pistil. x30.(L) Fl-54 mutant flower. 24L. p, pollen grains. Arrowsshow stamens without anther sacs. X18.(M) Fl-54 mutant inflorescence. 24L. /, flowers; fi,filaments; se, sepal-like structures with a stalk. x6.(N) Fl-82 mutant flower. 24L. xl8.(O) Fl-89 mutant flower. 24L. Arrows indicate horn-likeprojections. x22.(P) Transverse section of a pistil of the Fl-89 mutant, ov,ovules. Arrows show unfused septa.

11Y2M31/4

vb

db

Flower mutants of Arabidopsis 199

inside the tissue. On the other hand, the shape of thesepals rather resembles that of the leaves. The sepalsare flat and do not cover the young buds. There aretrichomes on both inner and outer surfaces of thesepals (Fig. 2E). As mentioned earlier, the normalsepals of the wild type do not have trichomes on then-inner surfaces, while the leaves do carry trichomes.The number and structure of other floral organs aregenerally normal (Table 1). Some mutant flowers,however, have four or five stamens (Fig. 31) morefrequently than the wild type.

Although the phenotypes of Fl-40 and Fl-48 mu-tants and that of the ap-2 mutant are generally quitedifferent from each other, similar phenotypes areoccasionally observed. In one specimen among thetwenty ap-2 mutant flowers examined, sepals hadstigmatic papillae at their apices and a few ovule-likeparticles at the margins just like the converted sepalsof the Fl-40 and Fl-48 mutants. Under the short-daygrowth conditions, some flowers of the Fl-40 and Fl-48 mutants had an extra pistil between the sepals andthe stamens on a short stem developed from the floralaxis (receptacle). In other Fl-40 and Fl-48 mutantflowers, two axillary flowers, each containing a pistiland stamens are formed on a short stem at a medianposition with respect to the inflorescence axis(Figs 2J, 3F). This phenomenon is known as ecblast-esis. In these flowers, the sepals are large and flat andhave trichomes on both inner and outer surfaces likethe sepals of the ap-2 mutant. Similar extra pistils areoccasionally found in the flowers of ap-2 mutantgrown under the short-day conditions (Figs 2K, 3J).These results show that the phenotypes of the threehomeotic mutants, Fl-40, Fl-48 and ap-2, becomemore similar to each other when grown under theshort-day conditions. The reason for this is notunderstood.

Flower structure and genetics of the Fl-54 mutantAnother mutant, Fl-54, shows variable phenotypes inthe shape, number and position of the floral organs(Figs 2L, 3K-N, Table 1). Several kinds of homeoticchanges are observed in the floral organs. Of the 42flowers examined, everyone had a somewhat differ-ent phenotype.

Most of the stamens lack anther sacs at the end ofthe filaments (Figs 2L, 3K). In some flowers, the topof the filaments is white and flat like a petal (Fig. 3N).In some other flowers, filaments are green, and havestigmatic papillae at the top, indicating partial con-version to carpels. Some stamens are extremely thinand are attached to the ovary wall. The number ofpetals is reduced to one or two (Table 1). Pollengrains are occasionally seen along the margins of thepetals (Figs 2L, 3M), indicating incomplete conver-sion of the petals to stamens. In some other flowers,

petals are green and look like sepals. Most of theflowers have three sepals, which are slender and haveno, or only a few, trichomes. In some flowers, themargins of the sepals are white, showing partialconversion to petals. It appears that the pistils are lessaffected by the mutation and set seeds well by meansof artificial pollination.

In flowers lacking some floral organs, the remain-ing organs are not necessarily positioned symmetri-cally as shown schematically in Fig. 3K-N. In somecases, short filaments and/or small knobs are visibleon the flower receptacles, at positions correspondingto the missing organs. These structures might be thefloral organs whose development had been abolishedin earlier stages.

The structure of the inflorescence is also abnormalin the Fl-54 mutant. It is known that the inflorescenceof Brassicaceae is racemose (Cronquist, 1981) and theflowers are formed along the inflorescence axis in aspiral arrangement at more or less constant intervals.In the Fl-54 mutant, however, the flowers are clus-tered in relatively short segments along the inflor-escence. Between the clusters of flowers are segmentsin which short green filaments and some sepal-likestructures with a short stalk are clustered (Figs 2M,3O). The sepal-like structures bend with respect tothe inflorescence axis. The shape and orientation ofthe structures are similar to those of the abaxial sepal(Fig. 1). When grown under the short-day conditions,two thin filaments are generated additionally from thejunction of the sepal-like structure and the stalk (datanot shown).

Several genetic crosses were performed to examinethe mutant gene of Fl-54. The mutation responsiblefor Fl-54 is nuclear, single and recessive. Allelismtests show that Fl-54 is not allelic with the ap-1, ap-2,clv-1, pi, Fl-40 or Fl-82 mutants (Table 2).

Flower structure and genetics of the Fl-82 mutantThe flowers of the Fl-82 mutant have extraordinarilylarge pistils (Fig. 2N), usually consisting of three ormore carpels (Fig. 3P-S, Table 1). In some flowers,the carpels are not fused well (Fig. 3Q). Anthers withpollen grains are sometimes attached at the side edgeof the unfused carpels. In some other flowers, thereare two separate pistils (Fig. 3R). Fertilization fre-quency of this mutant is generally low and the pistilsof unfused carpels are completely sterile. The numberand shape of other floral organs are usually normal,except that some mutant flowers occasionally havefive sepals, five petals and eight stamens (Fig. 3S). Itis likely, therefore, that the extra carpels are notgenerated from other organs by homeotic conversion,but rather they are formed by multiplication of thecarpel primordia.

Genetic analyses show that the Fl-82 mutation(s) is

200 M. K. Komaki and others

A sepalconvertedto a carpel

(Do

Marginal Centralbundle „. bundle

Twoadventitious

flowers

Ovules

Stamen lackinganther

Normal"v stamen

r-L <O,

,s/p

Normalsepal '-•

Petalsconvertedto stamens

Sepalsconvertedto leaves

oo /

Pollengrains

Long whiter—'• filament

Verythin

filaments

M ooo

5 sepals5 petals

8 stamens

A smallpistil

proliferated

O

» : Inflorescence axis>=? : Sepal*a^ : Converted sepal carrying ovules^ : Petal

O•

(1)

: Stamen' Filament: Pistil (two fused carpels): Two open carpels

Flower mutants o/Arabidopsis 201

Table 2. Genetic analyses of the mutant strains of Arabidopsis thaliana

StrainsNature ofmutation

Number ofloci Remarks

Fl-40

Fl-48

Fl-54

Fl-82

Fl-89

nuclearrecessivenuclearrecessivenuclearrecessivenuclearrecessivenuclearrecessive

single

single

single

not determined

single

allelic with ap-2 and Fl-48not allelic with ap-1, clv-1, pi, Fl-54, Fl-82 or Fl-89allelic with ap-2 and Fl-40not allelic with ap-1, clv-1, pi, Fl-54, Fl-82 or Fl-89not allelic with ap-1, ap-2, clv-1, pi, Fl-40 or Fl-82male sterile*not allelic with ap-1, ap-2, clv-1, pi, Fl-40 or Fl-54female semi-sterilenot allelic with ap-1, ap-2, clv-1 or Fl-40female sterilet

* Most of the anthers do not bear enough pollen.tLess than 1 % of the pistils set seeds in artificial pollination using wild-type pollen.

recessive and nuclear (Table 2). It is not clear,however, whether the mutation is single. The hetero-zygotes with the ap-1, ap-2, clv-1, pi, Fl-40 or Fl-54mutations display the normal phenotype, indicatingthat these mutant loci are not allelic with the Fl-82mutation(s). The phenotype of the clv-1 mutant issomewhat similar to that of the Fl-82 mutant. Thepistil of the clv-1 mutant consists of three or fourcarpels, which are always fused completely to form a

Fig. 3. Phenotypic variations of the mutant flowers.Flower structures and its diagram are shown.(A-F) Fl-40 and Fl-48 mutant flowers. Hatched crescentin the diagram indicates the converted sepals on themedian plane. (D,E) Fl-40 and Fl-48 mutant flowers withunfused carpels; (F) a flower with two adventitiousflowers.(G) A theoretical primitive carpel (redrawn from Eames,1931).(H-J) Apetala-2 mutant flowers. (H) A flower of normalarrangement. Each floral organ arises at thecorresponding positions to those of wild-type flowers.(I) A flower with four stamens; (J) A flower with a smallextra pistil.(K-N) Fl-54 mutant flowers. Number of floral organs isreduced. Stamens usually lack anthers. Position of organsis distorted in every flower. (K) Six filaments (maycorrespond to stamens) in the distorted positions; (L) aflower having a normal stamen; (M) a petal partiallyconverted to a stamen; (N) a long white filament in theposition of a petal. Very thin filaments present.(O) A Fl-54 mutant inflorescence. Thin filaments andsepal-like structures appear in the positions of flowers.(P-R) Pistils of Fl-82 mutant flowers. (P) Four fusedcarpels; (Q) three unfused carpels; (R) two separatepistils.(S) A Fl-82 mutant flower with five sepals, five petals andeight stamens.(T,U) Pistils of Fl-89 mutant flowers. (T) A pistil withtwo stigmas and horn-like projections; (U) a pistil withtwo stigmas but without clear horn-like projections.

club-like pod (McKelvie, 1962; Koornneef et al. 1983;Pruitt et al. 1987; Haughn & Somerville, 1988).

Flower structure and genetics of the Fl-89 mutantThe flower of the Fl-89 mutant has a pistil with twoclumps of stigmatic papillae and two horn-shapedgreen projections at the top (Figs2O, 3T,U). Thehorns are located at the top of the carpels, whereasthe clumps of stigmatic papillae reside on the seam-line of the carpels. Structural abnormalities of thismutant are also found inside the pistils. Although thecarpels are always fused completely and the ovulesseem to be normal, the septum tissues are not fused toform the partition (Fig. 2P). The number of the otherfloral organs is seemingly normal, but the sepals andthe petals are slightly more slender than those of thewild type.

This mutant grows slower than the wild type. Ittakes about 2 months for the first flowers to appear. Itremains to be clarified whether this delay is caused bythe same mutation responsible for the defects in floralstructure.

The results of the genetic analyses are listed inTable 2. The mutation is single, recessive and locatedon the nuclear genome. The allelism test shows that itis not allelic with any of the ap-1, ap-2, clv-1 and Fl-40mutations.

Discussion

The development, and morphology of the flowers ofArabidopsis thaliana have been studied geneticallyand many mutants having a distorted flower mor-phology have been isolated. The analyses of themutants clearly show that the developmental processas well as the morphology of the flowers are under thecontrol of numerous genes. When one of the genesmutates, drastic alterations may occur in the flower-ing process. According to a classical view, a flower is

202 M. K. Komaki and others

considered to be a modified shoot and floral organsare modified leaves developed from the primordialorgan (Esau, 1977). If a certain genetic regulatorysystem that determines the type of floral primordia isdisturbed by mutation(s), homeotic conversions mayoccur.

Of the mutants isolated, five mutants have beenextensively characterized: three of them (Fl-40, Fl-48,Fl-54) have homeotic variations accompanied by thedisappearance of several floral organs, and the rest(Fl-82, Fl-89) have alterations localized in the pistils.The three homeotic mutants have variable pheno-types, while the phenotypes are not so variable in thetwo pistil mutants. It is likely that the difference invariability among the mutant phenotypes is due to thenature of the functions of the mutant genes. It ispossible that the products of the homeotic genes maybe required throughout the process of flower develop-ment. Alternatively, these genes may work only at anearly stage (s) of flower development when primordialorgans are formed. In any case such homeotic genesmay cause more pleiotropic effects than other genesresponsible for flower development. In contrast, inthe pistil mutants, the mutated genes may work incells forming the primordial pistil (pistil primordia).Since the pistil is considered to be developed laterthan other organs, the mutations may not affect thedevelopment of other floral organs.

In the Fl-40 and Fl-48 mutants, the positions of thefloral organs affected by the mutations are mainlymedian to the inflorescence axis. As shown inFig. 3A-E, the converted sepals are usually locatedon the median plane, whereas the sepals of the wild-type are on the transverse plane. In flowers havingonly one or two stamens, the remained stamens are inthe transverse position. Additional pistils and axillaryflowers which sometimes appear under the short-daygrowth conditions are located on the median planebetween the sepals and the stamens (Fig. 3F). Thesepositional preferences of the mutational defectswould reflect the median distribution of the mutatedgene function in flower development.

Although the mutated gene of the Fl-40 and Fl-48strains is allelic with that of the ap-2 mutant, pheno-types of the former mutants are considerably differ-ent from that of the latter. In general, morphologicalaberrations are more profound in the Fl-40 and Fl-48mutants than in the ap-2 mutant. In view of the factthat the ap-2 mutation is dominant over the Fl-40 orFl-48 mutations, we assume that the genes in the Fl-40and Fl-48 mutants are more extensively damaged thanthose in the ap-2 mutant. The nature of these mu-tations will be unravelled when their genes are clonedand characterized. The other independently isolatedfloral mutants (flo2,flo3,flo4) (Haughn & Somerville,1988) appear to have phenotypes that are somewhat

similar, in some aspects, to those of the Fl-40 and Fl-48 mutants. It remains to be clarified whether thesemutations are allelic with each other or not.

The floral structure of the Fl-54 mutant is highlyvariable. Several homeotic variations were detectedin the sepals, petals and stamens. Some floral organsare missing or not fully developed. The positions ofthese floral organs often deviate from their standardpositions. The highly pleiotropic effects of this mu-tation probably indicate that the mutated gene prod-uct plays a crucial role in the development of thesefloral organs.

Of particular interest is the fact that the mutation inthe Fl-54 mutant results in the abnormal structures onthe inflorescence. Clusters of thin filaments and sepal-like structures are formed on the inflorescence axisbetween clusters of flowers. The position of the sepal-like structures relative to the axis suggests that theycorrespond to the abaxial sepal which appears first inflower development of Brassicaceae (Sattler, 1973). Itis likely, therefore, that the sepal-like structuresrepresent traces of aberrantly developed flowers inwhich only the abaxial sepal is developed but notother floral organs. If this is true, it would imply thatthe Fl-54 mutation may cause two different kinds ofdisturbance in the normal process of flower develop-ment. One is to stop the floral bud developmentimmediately after the appearance of abaxial sepals,and the other is that flowers are ultimately developedand pistils are formed but the flowers show homeoticchanges among floral organs. These two modes ofdisturbance somehow appear alternately on the in-florescence axis, as the inflorescence grows. Furtherstudy is necessary to clarify the mechanism underly-ing this phenomenon.

Two mutants, Fl-89 and Fl-82, have structuralabnormalities in the pistils. The stigma and septum ofthe pistils in the Fl-89 mutant are altered. Accordingto the pistil development scheme in Brassicaceae, twosepta grown from the region between the two pla-cental bands meet in the centre and fuse with eachother (Sattler, 1973). The developmental step inwhich the septa fuse is apparently blocked in the Fl-89mutant. The abnormal stigma structure, the twoclumps of papillae and the horn-shaped projectionswould be the result of incomplete septal fusion. Themutated gene of the Fl-89 mutant may be required forthe development of the pistil primordium, especiallyin the stages of septal fusion. The organization of thepistils in the Fl-82 mutant is also incomplete. Usuallythe pistils consist of three or more carpels which areoften not fused. Consequently, the ovules are ex-posed to air and the septa are not well developed. Inaddition to the structural abnormalities, the pistils ofFl-82 and Fl-89 are functionally defective, as judgedby the low fertilization frequency of the pistil.

Flower mutants of Arabidopsis 203

Although it is not known which step(s) of fertilizationis blocked in each mutant, the structural defects ofthe pistils would be responsible for the sterility ofthese mutants.

Attempts to identify the genes responsible for thesemutations are in progress in our laboratory. When thegenes are identified, their structure and expressionmay be examined in order to provide vital clues forunderstanding the mechanism of flower developmentand morphogenesis in higher plants.

We thank Drs K. R. Kranz (AIS, Frankfurt am Main,FRG) and E. M. Meyerowitz (Caltech) for providing seeds,Dr C. R. Somerville (Michigan State University) for theprocedures of mutagenesis, cultivation and artificial polli-nation, Dr R. Kodama (National Institute for Basic Bi-ology, Japan) for teaching us the thin-section techniques.This work was supported in part by grants from the Ministryof Education, Science and Culture and by a fund from theCIBA-GEIGY Foundation for the Promotion of Science.

References

CRONQUIST, A. (1981). An Integrated System ofClassification of Flowering Plants, pp. 436—441 & pp.446-449. New York: Columbia Univ. Press.

EAMES, A. J. (1931). The vascular anatomy of the flowerwith refutation of the theory of carpel polymorphism.Am. J. Bot. 18, 147-188.

ESAU, K. (1977). The flower: structure and development.In Anatomy of Seed Plants, second ed., chap 20: pp.375-401. New York: John Wiley.

ESTELLE, M. A. & SOMERVILLE, C. R. (1986). Themutants of Arabidopsis. Trends in Genetics 2, 89-93.

HAUGHN, G. W. & SOMERVILLE, C. R. (1988). Geneticcontrol of morphogenesis in Arabidopsis. Devi Genet.9, 73-90.

KOORNNEEF, M . , DE B R U I N E , J. H . & GoETTSCH, P.(1980). A provisional map of chromosome 4 ofArabidopsis. Arabidopsis Inform. Serv. 17, 11-18.

KOORNNEEF, M . , VAN EDEN, J. , H A N H A R T , C. J . , S T A M ,

P., BRAAKSMA, F. J. & FEENSTRA, W. J. (1983).Linkage map of Arabidopsis thaliana. J. Hered. 74,265-272.

KOORNNEEF, M. (1987). Genetic Maps 1987. (ed. S. J.O'Brien), pp. 742-745. New York: Cold Spring HarborLaboratory.

LAWRENCE, G. H. (1951). Taxonomy of Vascular Plants.pp. 520-524. New York: MacMillan.

MCKELVTE, A. D. (1962). A list of mutant genes inArabidopsis thaliana (L.) Heynh. Rad. Bot. 1, 233-241.

MEYEROWITZ, E. M. (1987). Arabidopsis thaliana. A. Rev.Genet. 21, 93-111.

MEYEROWITZ, E. M. & PRUITT, R. E. (1985). Arabidopsisthaliana and plant molecular genetics. Science 229,1214-1218.

MULLER, A. (1961). Zur Charakterisierung der Bliltenund Infloreszenzen von Arabidopsis thaliana (L.)Heynh. Kulturpflanze 9, 364-393.

PRUITT, R. E., CHANG, C , PANG, P. P.-Y. &

MEYEROWITZ, E. M. (1987). Molecular genetics anddevelopment of Arabidopsis. In Genetic Regulation ofDevelopment, pp. 327-338. New York: Alan Liss.

REDEI, G. P. (1975). Arabidopsis as a genetic tool. A.Rev. Genet. 9, 111-127.

SATTLER, R. (1973). Organogenesis of Flowers. APhotographic Text-Atlas, pp. 68-71. Toronto &Buffalo: Univ. Toronto Press.

(Accepted 5 July 1988)