During the development of multicellular organismsmany cell types are produced. Depending on theirposition, each cell perceives different signals, respondsthrough intracellular signalling pathways and, even-tually, adopts a specific cell fate. Subsequent cell dif-ferentiation usually involves complex changes. Forexample, cells exit the mitotic cycle or enter cycles ofENDOREDUPLICATION, the cellular architecture alters tomeet the functional requirements of the respectivecell type, and the metabolism of the cell changesaccording to its function. Compared with animals,plant development faces additional constraintsbecause the rigid cell walls prevent any cell move-ment. In plants, a few single-celled Arabidopsisthaliana model systems — in particular root hairsand trichomes — have greatly improved our under-standing of the development of single cells.

The goal of this review is to summarize how thestudy of A. thaliana trichomes facilitates the under-standing of development at the single-cell level. A largenumber of mutants have been characterized thatenabled the identification of subsequent developmentalprocesses. These include the selection of trichomes in afield of epidermal cells, cell-fate determination, changesin the cell-cycle mode and cell-shape control. Thegenetic, molecular and cell-biological analysis of tri-chome development has revealed only a few trichome-specific processes, as most developmental steps involvethe regulation of general cellular machineries. Therefore,studying the trichome system has provided uniqueinsights into the function of transcription factors, the

microtubule and actin cytoskeleton, the cell cycle andcell-death control. The study of all the developmentalstages of a single cell is a first step towards an under-standing of how general cellular processes are inte-grated during development.

Steps in trichome developmentShoot epidermal hairs are known as trichomes, a termthat is derived from the Greek word for hairs, trichos.Trichomes are found in most plants and can compriseeither single or several cells and can be secretory glan-dular or nonglandular1,2. The functions that areascribed to trichomes range from protecting the plantagainst insect herbivores and UV light, to reducingtranspiration and increasing tolerance to freezing3,4.

Trichomes are an excellent model system becausethey are of epidermal origin and are therefore easilyaccessible. In addition, A. thaliana trichomes are notessential for the plant under laboratory conditions,which facilitates the isolation of trichome-specificmutants5,6. So far, most studies have been carried outon leaf trichomes (FIG. 1). At the base of young leaves,single cells that are spaced out at regular distances in anarea of apparently equivalent PROTODERMAL CELLS developinto trichomes7,8. Incipient trichomes stop mitotic celldivisions and initiate endoreduplication cycles. As aresult, the trichome cell increases in size and changes itsdirection of growth such that it grows perpendicular tothe leaf surface. Further growth is characterized by atotal of about four endoreduplication cycles that resultin a DNA content of 32C (1C is the DNA content of

PLANT TRICHOMES: A MODEL FORCELL DIFFERENTIATIONMartin Hülskamp

During the past few years, the focus in plant developmental biology has shifted from studying theorganization of the whole body or individual organs towards the behaviour of the smallest unit ofthe organism, the single cell. Plant leaf hairs, or trichomes, serve as an excellent model system tostudy all aspects of plant differentiation at the single-cell level, including the choice of cell fate,developmental control of the cell cycle, cell polarity and the control of cell shape.

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 5 | JUNE 2004 | 471

University of Köln, BotanicalInstitute III, Gyrhofstraße 15,50931 Köln, Germany.e-mail: martin.huelskamp@uni-koeln.dedoi:10.1038/nrm1404

ENDOREDUPLICATION

A modified cell cycle in whichDNA replication continues inthe absence of mitosis andcytokinesis.

PROTODERMAL CELL

A young epidermal cell that hasnot yet differentiated into aspecialized cell type.

R E V I E W S

P L A N T C E L L B I O LO G Y

© 2004 Nature Publishing Group

472 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/molcellbio

R E V I E W S

emerging picture is that only very few genes are, in fact,trichome specific. Most genes are relevant for many celltypes and are involved in general cellular processes (FIG. 1).It seems that mutations in these genes have little effect inmost cell types but are crucial during trichome develop-ment — possibly because trichomes, with their rapidgrowth and enormous size, are more demanding.

Trichome patterning and initiationWild-type trichomes are initiated with an average dis-tance of about three cells between developing trichomes,and almost never form directly next to each other — aswould be expected if they were randomly distributed —which indicates that there must be an underlying pat-terning mechanism7,8. A mechanism that wouldexplain trichome patterning by a standardized cell-division pattern that segregates trichome cell fates wasexcluded by clonal analysis8,11. It is therefore hypothe-sized that trichome selection is based on a mutual-inhibition mechanism12–15 (FIG. 2): cells that are initiallyequivalent produce a trichome-promoting factor (orfactors) that activates a factor (or factors) that sup-presses trichome development in the neighbouring cells.This way, cells begin to compete and, due to stochastic

the unreplicated haploid genome), which is accompa-nied by rapid cell enlargement7,9. The growing cellundergoes two consecutive branching events, the orien-tations of which are co-aligned with respect to thebasal–distal leaf axis10.

A large number of mutations that affect trichomedevelopment were identified in several mutagenesisscreens5,6. The mutations helped to define regulatoryprocesses in trichome development according to theirspecific developmental defects (FIG. 1). The selection of tri-chome cells and the initiation of the trichome cell fate areunder the control of a small group of so-called patterninggenes. One gene seems to specifically translate the pat-terning cues into cell-fate differentiation. The switch frommitotic cycles to endoreduplication cycles and the num-ber of endoreduplication cycles are controlled by theendoreduplication genes. A large number of genes areknown to be important for branching. The directionalityof expansion growth is affected in the so-called distortedmutants. One mutant is known to cause unscheduled celldeath, and several other mutants seem to affect the matu-ration of the trichome.

Now that the genetic interactions are well under-stood and most of the genes have been cloned, the

Figure 1 | Trichome development. A schematic presentation of trichome development is shown at the top. Below, representativeexamples of mutant or overexpression phenotypes, which enabled the identification of the developmental processes, are shown.First, a trichome cell is selected from protodermal cells. The try cpc (triptychon caprice) double mutant shows a defective trichomepattern. A cell that has adopted the trichome cell fate switches from mitotic cycles to endoreduplication cycles. A situation in whichthis switch is suppressed is exemplified by a trichome in which a specific B-type cyclin, CYCB1;2, is overexpressed under thecontrol of the trichome-specific GL2 promoter (GL2:CYCB1;2). The glabra2 (gl2) mutant phenotype indicates that cell-fatedetermination requires specific gene functions. The angustifolia (an) mutant illustrates the requirement of genes to undergo properbranching. The directionality of expansion growth requires a group of several so-called DISTORTED genes, of which the wurm (wrm)mutant is shown. The kaktus (kak) mutant has twice the DNA content of wild-type plants and represents the class of mutants thatregulates the number of endoreduplication cycles. Cell-death control can also be studied using trichomes. Certain mutants — andthe overexpression of the cell-cycle kinase inhibitor/interactor of cyclin-dependent kinases (ICK) under the control of the trichome-specific GL2 promoter (GL2:ICK), which is shown here — lead to unscheduled cell death. Finally, several mutants have a fragile andglassy appearance and are therefore thought to be involved in the maturation of the trichome cell (not shown). The drawings ofdevelopmental stages are adapted from REF. 6 and the gl2 image is reproduced from REF. 98. Other images are reproduced withpermission as follows: try cpc from REF. 19 © (2002) Macmillan Magazines Ltd; GL2:CYCB1;2 from REF. 35 © (2002) Elsevier; an fromREF. 10 © (1997) The Company of Biologists Ltd; wrm from REF. 76 © (2003) Springer–Verlag GmbH; kak from REF. 7 © (1994)Elsevier; GL2:ICK from REF. 91 © (2003) The American Society of Plant Biologists.

try cpc GL2:CYCB1;2 gl2 an wrm kak GL2:ICK

Trichome selectionSwitch from mitosisto endoreduplication Cell-fate determination Branching Expansion growth

Regulation of the number of endoreduplication cycles Cell-death control

© 2004 Nature Publishing Group

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 5 | JUNE 2004 | 473

R E V I E W S

function as positive regulators of trichome develop-ment. Mutations in the GLABRA1 (GL1) and TRANS-PARENT TESTA GLABRA1 (TTG1) genes each result inthe complete absence of trichomes16,17, whereasGLABRA3 (GL3) and ENHANCER OF GL3 (EGL3)function in a redundant manner — gl3 mutants exhibitfewer trichomes compared with wild-type plants,whereas gl3 egl3 double mutants are devoid of tri-chomes18. The trichome-suppressing genes are repre-sented by three redundantly acting genes (see below).Mutations in the TRIPTYCHON (TRY) gene result intrichome clusters, mutations in the CAPRICE (CPC)gene cause an increased number of trichomes19,20 and asingle mutant of ENHANCER CAPRICE TRIPTY-CHON1 (ETC1), which is an enhancer of cpc and trymutants, is indistinguishable from wild-type plants21.

Genetic analysis has established the functional rela-tionships between the four positive factors. The findingsthat co-overexpression of GL3 and EGL3, as well as GL3together with GL1, can rescue the ttg1-mutant pheno-type indicates that TTG1 functions upstream of thesegenes and that the other three factors function togetherat the same point in the pathway18,22. These data havebeen confirmed at the molecular level.All trichome-pro-moting genes, except for TTG1, encode putative tran-scription factors. GL1 encodes a MYB-RELATED TRANSCRIPTION

FACTOR23, GL3 a BASIC HELIX–LOOP–HELIX (BHLH) FACTOR22, EGL3is a close homologue of GL3 (REF. 18), whereas TTG1encodes a WD40 PROTEIN whose molecular function isunknown24.Yeast two-hybrid data indicate that the fourpositive factors form a transcriptional-activation com-plex in which GL3 forms a homodimer that binds toGL1 (REF. 18,22). The GL3 protein also binds to the TTG1protein, but through a different domain. No directinteraction was found between GL1 and TTG1 (REF. 22).It is likely that GL1 and GL3 mediate the transcriptionalactivation, as both proteins contain transcriptional-acti-vation domains. This complex is expected to be active intrichome precursor cells and be inactivated in all otherprotodermis cells by one or more known negative regu-lators of trichome initiation.

The negative regulators TRY, CPC and ETC1 allbelong to a small family of single-repeat MYB proteinswith no obvious transcriptional-activation domain19–21.Overexpression of any of these proteins abolishes

fluctuations, individual cells will gain higher levels of thepromoting factor, produce more of the suppressing fac-tor and, in turn, inhibit the neighbouring cells morestrongly. Eventually, these cells would become commit-ted to the trichome cell fate. For this mechanism towork a number of criteria have to be met (for theoreti-cal considerations, see BOX 1). First, the positive and thenegative regulators need to be involved in a feedbackloop with the activator activating the inhibitor and theinhibitor inhibiting the activator. A second requirementis that the inhibitor can move.

The genetic and molecular analysis of the trichome-patterning genes is consistent with this model, althoughlittle proof is available that could directly demonstratethe patterning mechanism. Both the positive and thenegative regulator are represented by a group of severalfactors (FIG. 2). Four of the trichome-patterning genes

MYB-RELATED TRANSCRIPTION

FACTOR

A transcription factor thatcontains a DNA-bindingdomain that shows sequencesimilarity to vMYB, the first-described member of this family.

BASIC HELIX–LOOP–HELIX

(BHLH) FACTOR

A protein that contains two α-helices separated by a loop(the HLH domain), which bindsDNA in a sequence-specificmanner.

WD40 PROTEIN

A 40-amino-acid-long proteinmotif that contains a WDdipeptide at its carboxyterminus. This domain is found in many functionallydiverse proteins and mediatesprotein–protein interactions.

Figure 2 | Redundancy of trichome-patterning genes. The top row of phenotypes shows theredundancy of the negative regulators of trichome patterning. From left to right, wild-type leaves,the patterning defects of triptychon (try) single, try caprice (cpc) double and try cpc enhancercaprice triptychon (etc1) triple mutants are shown. Compared with wild-type plants, try mutants have only a few trichome clusters that contain 2–3 trichomes. Theadditional removal of CPC results in an increased cluster size (up to 40 trichomes) and mutationsin ETC1 lead to the formation of trichome clusters with several hundred trichomes. It isnoteworthy that cpc single mutants (not shown) show a subtle increase of the trichome densityand that etc1 mutants show no phenotype. The lower row of phenotypes shows that GLABRA3(GL3) and ENHANCER OF GL3 (EGL3) are redundant. Both single mutants show a slightreduction in trichome number compared with wild-type plants and the double mutant lackstrichomes. The images are reproduced with permission as follows: top row from REF. 21 © (2004)Elsevier; lower row from REF. 18 © (2003) The Company of Biologists Ltd.

WT try try cpc try cpc etc1

WT gl3 egl3 gl3 egl3

Box 1 | Theoretical model to explain two-dimensional patterning

A theoretical model that explains how a de novospacing pattern can be established, which issimilar to that observed for trichomes, wasformulated by Meinhardt and Gierer95. Theessence of this model is that an activator controlsthe production of its own inhibitor, which canfunction over a long distance. This, however, isnot sufficient to create the initial differences in a uniform field of cells. For this, it is necessary that the activator has self-enhancing properties to form a so-called autocatalytic loop. This allows the rapid amplification of small differences thatare caused by the stochastic fluctuation of biomolecules to trigger further patterning. One non-intuitive prediction of thismodel is that both the activator and the inhibitor show the highest expression at the same locations and not, as might beassumed, at mutually exclusive locations. Three steps of a simulation of this model that results in a spacing pattern areshown (see online supplementary information S1 (movie); courtesy of Hans Meinhardt).

© 2004 Nature Publishing Group

474 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/molcellbio

R E V I E W S

Trichome differentiationThe homeodomain leucine-zipper protein that isencoded by the GL2 gene29,30 is thought to translate thecues that are provided by the patterning genes into cell-specific differentiation of several epidermal cell typesincluding the seed coat, root hairs and trichomes17,29,30

(BOX 2). This is documented by the finding that co-overexpression of GL1 and the maize R gene product(a homologue of GL3) results in an increased andectopic expression of GL2 (REF. 31). Although some evi-dence indicates that GL2 might also have a role in tri-chome patterning, most of the available data indicatethat GL2 triggers downstream differentiation events32.The gl2-mutant-trichome phenotype is characterizedby undifferentiated trichomes that resemble the com-bined phenotypes of various trichome-morphogenesisgenes, which indicates that GL2 activates trichome-spe-cific differentiation genes7,29. Supporting evidence comesfrom the root-hair system, where it was shown that GL2regulates the gene that encodes phospholipase Dζ1,which, in turn, promotes root-hair differentiation33.

Cell-cycle control during trichome developmentTrichome cell-cycle-regulation mutants affect either theswitch from mitotic cycles to endoreduplication cycles,or the number of endoreduplication rounds, andthereby the ploidy level (FIG. 3).

trichome formation. They seem to function in a highlyredundant manner: try mutants have small trichomeclusters consisting of two or three trichomes, try cpcdouble mutants have large clusters of up to 40 trichomesand in try cpc etc1 triple mutants, fields of several hun-dred trichomes are observed (FIG. 2)7,19,21. These negativefactors seem to interfere with the function of the tran-scriptional-activation complex by a competition mech-anism25. Three-hybrid analysis has shown that theinteraction between GL1 and GL3 is counteracted byTRY, thereby disturbing the formation of the proposedfunctional transcriptional-activation complex26.

How cell–cell communication and a regulatoryfeedback loop are achieved is, at present, unknown.However, evidence is available for root-hair patterningin A. thaliana, which requires a set of identical andclosely related genes (BOX 2). Here it was shown — usinga green fluorescent protein (GFP)–CPC fusion protein— that the negative regulator CPC moves betweencells, probably via PLASMODESMATA, which indicates thattravelling transcription factors can mediate cell–cellcommunication27. It was also shown that the positiveroot-hair-patterning genes WEREWOLF (WER; a GL1homologue) and GLABRA2 (GL2) are involved in anegative-feedback loop with CPC 28. It is conceivablethat a similar mechanism operates during trichomepatterning (BOX 2).

CORTEX CELL

The tissue between the vascularbundle and the epidermis. In A. thaliana, this is a single celllayer.

PLASMODESMATA

Cell–cell connections in plantsthrough which macromolecules,including RNA and proteins, canbe transported in a regulatedmanner.

Box 2 | Comparison of trichome and root-hair patterning

Root hairs seem to be specified by position-dependent induction as root hairs are only formed over the cleft between twounderlying CORTEX CELLS (see figure, part a). Epidermal cells that are positioned on a cortex cell do not form root hairs.Most of the genes (or their homologues) that are involved in root-hair patterning are also involved in trichomedevelopment (see figure, part b). The proposed models for trichome and root-hair patterning are similar, but theresulting phenotypes are different. Whereas those cells that eventually express GLABRA2 (GL2) in the shoot becometrichomes, GL2-expressing root epidermal cells adopt a non-root-hair cell fate. For both systems it is postulated that inthose cells that express the GL2 gene, a trimeric active complex is formed that consists of a MYB-related transcriptionfactor (GLABRA1 (GL1) in trichomes or WEREWOLF (WER) in roots), a basic helix–loop–helix protein (GLABRA3(GL3) and the homologous ENHANCER OF GL3 (EGL3)) and a WD40 protein (TRANSPARENT TESTA GLABRA1(TTG1)). This active complex is thought to activate GL2 and the expression of the negative regulators TRIPTYCHON(TRY), CAPRICE (CPC) and CAPRICE TRIPTYCHON1 (ETC1). These travel into the neighbouring cells where theycompete with the MYB-related transcrption factor for binding to the complex, which causes the complex to becomeinactivated. Identical or homologous proteins are shown in the same colour.

TTG1WER WERTTG1

GL1GL1

TTG1

TRY,CPC,ETC1

TRY,CPC,ETC1

TRY,CPC,ETC1

TRY,CPC,ETC1

TRY,CPC,ETC1

GL2

GL2 GL2

Trichome development No trichome development

No root-hair development Root-hair development

GL3,EGL3

GL3,EGL3 GL3,

EGL3 TTG1

TRY,CPC,ETC1

GL2

GL3,EGL3

Cortex cell Cortex cell

b

a

© 2004 Nature Publishing Group

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 5 | JUNE 2004 | 475

R E V I E W S

mutants. By contrast, the initiation of mitotic cycles insim mutants is independent of D-type cyclins36.

The number of trichome endoreduplication cycles isaffected in at least ten mutants that display either loweror higher ploidy levels than normal. The picture that isemerging from their genetic and molecular analysisis that several different molecular pathways are involvedin the regulation of trichome endoreduplication cycles.

Regulation by patterning genes. Two of the patterninggenes that are described above, GL3 and TRY, alsofunction as positive and negative regulators ofendoreduplication cycles; try trichomes have a DNAcontent of 64C and different gl3 alleles exist that haveeither a reduced or an increased DNA content7,26. Thisdual function raises the fascinating possibility thattrichome cell-fate choice is functionally linked withcell-cycle regulation.

DNA-catenation-dependent endoreduplication. ROOTHAIRLESS2 (RHL2) and HYPOCOTYL6 (HYP6) arepositive regulators of endoreduplication cycles in tri-chomes and in other cell types37. Both are plant homo-logues of the archaeal DNA TOPOISOMERASE VI complex.These topoisomerases can DECATENATE DNA and pro-mote ATP-dependent separation of entangled DNA37.It is unclear whether the observed defect in the progres-sion of trichome endoreduplication is due to a physicalblock of further DNA replication or the activation of acheckpoint that controls progression of the endoredu-plication cycle.

Regulation by plant hormones. The plant hormonegibberellin promotes endoreduplication cycles. Inspindly (spy) mutants, which exhibit a constitutive gib-berellin response, trichomes have a DNA content of64C (REF. 38). Conversely, in a mutant that is incapableof gibberellin synthesis, ga1-3, no trichomes areformed39,40.

Regulation by protein degradation. A class of four tri-chome mutants, kaktus (kak), rastafari (rfi), poly-chome (poc) and hirsute (hir), all show a very similarphenotype: they all have a ploidy level of 64C. Thecloning of the KAK gene revealed that it encodes aprotein with sequence similarity to a UBIQUITIN E3

LIGASE41,42. It is therefore assumed that ubiquitin-regu-lated protein degradation controls the progression ofendoreduplication.

Regulation by a cell-death pathway. Two lines of evi-dence indicate that a pathway exists that controls boththe progression of endoreduplication cycles (andmitotic cycles) as well as programmed cell death. First,the CONSTITUTIVE PATHOGEN RESPONSE5(CPR5) gene is involved in both processes. Second,overexpression of an inhibitor of the cell-cycle kinaseINHIBITOR/INTERACTOR OF CYCLIN-DEPEN-DENT KINASES/KIP-RELATED PROTEINS(ICK/KRP) leads to reduced ploidy and early tri-chome cell death (see below).

The SIAMESE (SIM) gene suppresses the switchfrom mitotic divisions to endoreduplication cycles34.In sim mutants, trichomes are multicellular and con-tain between 2 and 15 cells. If the first cycle is alreadymitotic, two trichomes are formed instead of one; ifthe switch to endoreduplication from mitotic divisionsis late, multicellular trichomes that are morphologi-cally normal are formed. As SIM has not been clonedyet, little is known about the molecular mechanismthat is involved. However, some information is avail-able from a different line of experiments in which therole of known cell-cycle genes in the control ofendoreduplication was tested. The expression of cell-cycle genes that are normally not active in trichomeswas used to test the effects on trichome endoreduplica-tion. To avoid organism-wide defects that could causesickness or even lethality, trichome-specific gene pro-moters were used. B-TYPE AND D-TYPE CYCLINS could triggerthe formation of multicellular trichomes, and B-typecyclins are involved in the transition from G2 phase tomitosis. The overexpression of a specific B-type cyclin,CYCB1;2, is sufficient to switch from endoreduplica-tion cycles to mitotic cycles, which indicates that B-typecyclins are important for this regulatory step35.Surprisingly, the overexpression of the D-type cyclinCYCD3;1 can also lead to the switch36. D-type cyclinsare thought to control the transition from the G1 tothe S phase of the cell cycle in animals; but these resultsindicate that, in plants, they have an additional func-tion in regulating the entry into mitosis. In simmutants, CYCB1;2 is expressed in trichomes, whichindicates that SIM inhibits the expression of mitoticcyclins. However, this is not its only function, as over-expression of CYCB1;2 in a sim mutant backgroundshows a much stronger phenotype than the single

B-TYPE AND D-TYPE CYCLINS

Cyclins regulate cell-cycleprogression throughinteractions with cyclin-dependent protein kinases.B-type cyclins regulate entryinto mitosis, whereas D-typecyclins are important in G1phase and, in plants, also forentry into mitosis.

DNA TOPOISOMERASE

An enzyme that can cleave andreligate the DNA to allow a morerelaxed DNA configuration.

DECATENATE

During DNA replication, sisterduplex molecules becomeinterlinked (catenated).Decatenation is the separation oftwo entangled chromosomes.

UBIQUITIN E3 LIGASE

An enzyme that attachesubiquitin to a protein, therebymarking it for degradation in theproteasome.

Figure 3 | Regulation of the cell cycle in trichomes. The mitotic cell cycle proceedsthrough four phases: M (mitosis), G1 (gap 1), S (synthesis) and G2. Endoreduplication cyclescut the cell cycle short by skipping the G2 and M phases. Processes that affect the numberof endoreduplication cycles (blue) were identified by mutant analysis. The factors that areinvolved in the switch from mitosis to endoreduplication are indicated in red. Overexpressionstudies have shown that the mitotic cyclin CYCB1;2 is sufficient to trigger complete celldivisions. The SIAMESE (SIM) gene represses CYCB1;2 as sim mutants express CYCB1;2 intrichomes. Also CYCD3;1 overexpression causes a switch from endoreduplication to mitosis.In contrast to the situation with CYCB1;2, however, the total number ofmitotic/endoreduplication cycles is drastically increased during CYCD3;1 overexpression,which indicates an additional role for this cyclin at G1.

M

S

G1

G2Ubiquitin-dependent protein degradation

Patterning

DNA catenation

Cell death

CYCD3;1

CYCB1;2

SIAMESE

Hormones (gibberellin)

Endoreduplication

© 2004 Nature Publishing Group

476 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/molcellbio

R E V I E W S

branches7,47. The reduction of the ploidy level results in areduced branch number — as found in cpr5, gl3 rhl2and hyp6 (REFS 7,37,48). It is likely that this observed con-trol of trichome branching by the ploidy level is indirect,and that the cell size or the time of actual cell growthprovides the frame for branch initiation.

Regulation by microtubules. Microtubules have animportant role in trichome branching, as shown inexperiments with microtubule antagonists. If the micro-tubule cytoskeleton is defective during trichome growth,the cell expands almost equally in all directions (this isknown as isotropic growth) and does not initiatebranches49. Several branching genes encode componentsthat are involved in the biogenesis of α/β-tubulindimers, the formation and stability of microtubules ormicrotubule-based transport processes. Two weakmutants of TUBULIN FOLDING COFACTOR (TFC)Aand TFCC — which are involved in the correct foldingof tubulins and therefore the formation of assembly-competent α/β tubulin dimers — exhibit a ‘bloated’ andunderbranched trichome phenotype50–52. The analysis ofthe microtubule cytoskeleton in these mutants shedssome light on how the microtubules are reoriented dur-ing branch initiation. As the microtubule density andorientation is normal, it is likely that the failure ofbranch formation is due to problems in de novo synthe-sis rather than in the reorientation of pre-existingmicrotubules. If branching requires the synthesis ofnew microtubules, it is conceivable that it also requiresthe fragmentation of pre-existing ones to allow a reori-entation of growth. This view is supported by thereduced-branching phenotype of mutants of theKATANIN-P60 gene53–55. Katanins are known to cut pre-existing microtubules into smaller fragments.Microtubule reorientation during branch formation istherefore assumed to be controlled by severing pre-existing microtubules combined with de novo synthesis.

The spatial control of the orientation of micro-tubules is regulated by at least two branching genes. TheFASS/TONNEAU2 gene regulates microtubules inthe context of cell divisions as can be inferred fromthe observation that in fass/tonneau2 mutants cell-divi-sion orientation is randomized56,57. It encodes a novelprotein, phosphatase-2A regulatory subunit, whichindicates that it regulates microtubules by proteinphosphorylation58. The second regulator of micro-tubule organisation is SPIKE59. This protein showssequence similarity to CDM-family adaptor proteins(Caenorhabditis elegans CED-5; Homo sapiensDOCK180; Drosophila melanogaster MYOBLASTCITY). These proteins function as guanine nucleotide-exchange factors (GEFs)60 and are thought to modulatethe cytoskeleton through small RHO-like GTPases(known as ROPs in plants)61.

In addition, specific microtubule-based transportprocesses seem to be important for branch formation.The branching gene ZWICHEL (ZWI) encodes acalmodulin-binding kinesin motor protein that bindsmicrotubules in a calmodulin-dependent manner62–67.The activity of ZWI is modulated by the KIC protein,

Trichome branchingThe typical three-dimensional branching pattern oftrichomes is a unique model system for studying howseveral axes of polarity and cell morphogenesis areestablished. Except for the sim mutant, all mutantsdescribed so far affect the number of branches butnot their orientation with respect to each other.Genetic and molecular data indicate that severalindependent molecular pathways participate in tri-chome branching7,10,43–46 (FIG. 4).

Regulation by endoreduplication levels. The number ofbranches on a trichome depends on the ploidy levelof the cell. Tetraploid plants, in which the DNA con-tent of all cells is doubled, have trichomes with super-numerary branches47. Similarly, mutants that havetrichomes with increased DNA levels — such as kak,poc, rfi, try and spy — have trichomes with up to eight

Figure 4 | Regulation of trichome branching. Trichome branching is controlled by at leastfour different molecular processes. The molecular analysis of the ANGUSTIFOLIA (AN) geneindicates that transcriptional regulation and/or Golgi-related processes are important forbranching. The corresponding an mutant is underbranched. Several mutants that affectmicrotubule function or organization have reduced branching; for example, the tubulin foldingcofactor c (tfcc) mutant has two short branches. The branch number also correlates with theploidy levels. Higher ploidy levels lead to more branches, and trichomes with reduced ploidylevels have fewer branches. The triptychon (try) mutant shown here has twice the DNA contentof wild-type cells and has five branches. A key regulator of branching is the STICHEL (STI)gene, as the corresponding sti mutant trichomes do not initiate branches. However, thebiochemical mechanisms through which STICHEL regulates branching are, at present,unknown. The images are reproduced with permission as follows: an, try and sti from REF. 10 ©(1997) The Company of Biologists Ltd; tfcc from REF. 51 © (2002) Elsevier; WT from REF. 76 ©(2003) Springer–Verlag GmbH.

Golgi/transcription Microtubules Endoreduplication STI dependent

WT

an tfcc try sti

© 2004 Nature Publishing Group

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 5 | JUNE 2004 | 477

R E V I E W S

all the available data by assuming that branching is evo-lutionarily derived from multicellular trichomes, inwhich branching is the result of a certain division pat-tern (BOX 3).

Directionality of trichome cell expansionLike most plant cells, trichomes enlarge several-foldduring the later stages of differentiation and expandin a polarized manner. This expansion occurs, unlike inthe growing tip of root hairs or pollen tubes, along thewhole cell surface75,76. The directionality of expansiongrowth is affected in mutants of eight genes, which arecollectively known as the DISTORTED genes. All dis-torted mutants show a very similar phenotype: tri-chomes show turns and twists, some regions of thecell become bulged and others are underdeveloped.Following the movement of small beads that had beenplaced on the trichome surface, it was shown that thisphenotype is caused by the regionally unbalancedexpansion of the cell76.

Findings from different experimental approachesindicate that the directionality of trichome cell expansiondepends on the actin cytoskeleton. First, the applicationof drugs that interfere with actin function perfectly phe-nocopies the distorted mutant phenotype75,77. Second,the actin cytoskeleton is organized aberrantly in dis-torted mutants (FIG. 5b)75–77. Third, all DISTORTED genesthat have been cloned so far encode components of theARP2/3 COMPLEX78–81, which promotes actin polymerizationby enhancing F-actin nucleation and side-binding activ-ities that result in the initiation of fine actin filamentsfrom pre-existing F-actin82,83.

The analysis of the distorted mutants demonstratedthat actin has a role in expansion growth that goesbeyond its mere requirement for general growth. Theobservation that in distorted mutants actin-based move-ment of organelles, such as peroxisomes or the Golgi, isnot generally impaired indicates that F-actin is stillfunctional (see supplementary information S2 (movie)and S3 (movie))78,79. Defects were found locally in thoseparts of the cell that were not growing (FIG. 5d). Non-growth regions contain heavily bundled actin, whereas

which binds to ZWI in a Ca2+-dependent manner68.This indicates that ZWI-dependent transportprocesses might ultimately be controlled by the intra-cellular second messenger Ca2+.

Regulation by transcription or Golgi-related processes.The ANGUSTIFOLIA (AN) gene regulates branchingby two possible pathways, by Golgi-related transportprocesses or by transcriptional co-activation. Itencodes a protein with sequence similarity to carboxy-terminal binding protein (CtBP) and brefeldin-A-ribosylated substrate (BARS)69,70. In D. melanogaster,CtBP binds to the zinc-finger transcription factors andfunctions as a transcriptional co-repressor71. In the rat,BARS proteins were identified as proteins that areADP-ribosylated after treatment with the fungal toxinbrefeldin A. Brefeldin-A treatments result in the trans-formation of Golgi stacks into a tubular-reticular net-work and it is therefore thought that BARS is involvedin Golgi functions72,73. Biochemical data are not avail-able for the plant CtBP/BARS protein; however, thefindings that an mutants have microtubule defects andthat AN physically interacts with ZWI in a yeast two-hybrid screen indicates that AN regulates microtubuleorganization69.

Regulation by the STICHEL gene. The STICHEL (STI)gene regulates trichome branching in a dosage-depen-dent manner; branch reduction is subtle in weak sti alle-les, becomes more pronounced in stronger alleles andtrichomes are unbranched in null-alleles. Conversely,overexpression of STI leads to extra branch formation74.This genetic behaviour indicates a key regulatory rolefor STI, although its molecular function is still elusive.STI encodes a protein that contains a domain withsequence similarity to eubacterial DNA-polymerase-IIIsubunits. However, it is unlikely that STI functions as aDNA polymerase subunit, as no replication effects werefound to be associated with the branching phenotype.

An underlying scheme of how branch formation iscontrolled is not evident from the current analysis of thebranching genes. One model, however, accommodates

PRE-PROPHASE BAND

A dense band of microtubules atthe cell cortex that appearsbefore the start of cell division inplants. Its position marks thefuture division plane.

PHRAGMOPLAST

A fibrous structure between thedaughter nuclei at telophase inplant cells; also known as the cellplate.

ARP2/3 COMPLEX

(Actin-related protein 2/3).A multi-protein complex thatconsists of seven differentproteins and initiates new actinfilaments on pre-existing ones.

Box 3 | The evolutionary origin of trichome branching

In plant species other than Arabidopsis thaliana,trichomes are frequently multicellular and branching isinitiated when cell division occurs in a plane that isperpendicular to the main direction of growth2 (seefigure). For example, the branches of the multicellularand branched hairs in Verbascum are initiated by celldivisions that are perpendicular to the main growth axis.The finding that a single mutation in the SIAMESE (SIM)gene is sufficient to produce multicellular trichomes in A. thaliana supports such an evolutionary origin34. According to this model, all aspects concerning the establishment ofthe orientation of the division plane and the corresponding cell polarity are still operating, even in the absence of theactual cell division43. Several observations support this model, including the correlation of ploidy level and branchnumber and the finding that the function of the kinesin motor protein ZWICHEL (ZWI) is linked to the cell cycle96,97.ZWI is localized to the PRE-PROPHASE BAND and the PHRAGMOPLAST, and the injection of ZWI antibodies into multi-cellularstamen hairs of Tradescantia virginiana results in a metaphase arrest and abnormal cell-plate formation96,97.Figure adapted from REF. 2.

© 2004 Nature Publishing Group

478 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/molcellbio

R E V I E W S

RHO and RAC/CDC42 signal-transduction pathwaysthat are known in animals and yeast are, in principle,present in plants, although they are strongly modified. Inagreement with this, ROPs were shown to control thelocal actin configuration in epidermal cells and down-stream components, such as the HSPC300 (haematopo-etic stem/progenitor-cell clone-300) complex, are knownto be involved in the control of actin organization84–86.

Cell-death control in trichomesThe analysis of trichome development has revealed twopathways that suppress cell death and also regulateendoreduplication (see above). One pathway is repre-sented by ICK/KRP, which shows homology to the ani-mal cell-cycle inhibitor p27Kip1 (REF. 87). In animals,p27Kip1 can induce apoptosis in the absence of growthfactors in some specific cell types88. When ubiquitouslyexpressed in the whole plant, ICK/KRP causes severegrowth reduction87,89,90, and when expressed under thecontrol of a trichome-specific promoter, trichomecells stop endoreduplication cycles after two cyclesand begin to die with symptoms that are characteristicof programmed cell death, such as the degeneration ofCHROMOCENTRES and nucleoli91.

A second pathway is linked to the response of plantsto plant pathogens. A number of mutants mimic theplant pathogen response. Many of these mutants show acell-death phenotype combined with growth defects92.One of these mutants, cpr5, shows a trichome pheno-type that is similar to that of ICK/KRP-overexpressinglines; trichomes have a ploidy level of about 8C andundergo unscheduled cell death48. It seems that in bothcases cell-cycle or endoreduplication-cycle progressionand the control of cell death are somehow linked; how-ever, the mechanistic basis of this link remains to bedetermined.

Control of maturationTrichome maturation is affected in a group of diversemutants in which adult trichomes appear transparentor underdeveloped. Three poorly characterizedmutants, chablis, chardonnay and retsina, have transpar-ent trichomes and the underdeveloped trichomemutant has no papilla on the trichome surface93. Thetrichome birefringence mutant is defective in the pro-duction of cellulose94.

Conclusions and perspectivesAlmost all trichome genes are involved not only intrichome development, but also in the developmentof other cell types and represent important compo-nents of generally important regulatory pathways.The analysis of trichome initiation has uncovered anevolutionarily conserved gene cassette of transcrip-tion factors that are involved in patterning processesand anthocyanin-synthesis control. Their evolutionand functional diversification will be very interestingto study. Also, the theoretical model that explainspattern formation (BOX 1) is far from being proven; forexample, at present, there are no target promotersknown that could be used to test the genetic predictions.

regions in the distorted mutants that exhibit growthcomprise a fine network of actin known as ‘fine actin’. Itis conceivable that the creation of a local fine-actin net-work promotes the transport of membrane and cell wallmaterial for the actual growth. It is speculated that theactin cytoskeleton is also involved in the fusion of mem-branes, as the fusion of small vacuoles, which normallyleads to the formation of the large central vacuole, doesnot take place in distorted mutants79.

It is unknown how the ARP2/3-complex-dependentformation of fine actin is spatially controlled in tri-chomes. Some of the canonical pathways such as the

CHROMOCENTRE

A region in plant chromosomesthat comprises heterochromatinand coincides with centromeresduring meiosis.

Figure 5 | Control of expansion polarity. Trichome cell expansion is controlled by the actincytoskeleton. a | In wild-type cells, actin is organized in long filaments (as indicated by the arrow).b | Mutants of the genes that are collectively known as the DISTORTED genes exhibit fragmentedactin (as indicated by the arrow). All distorted-gene mutants, except for one, show the samephenotype, which is a result of the fact that trichome expansion is no longer coordinated. c | Thephenotype of one distorted-gene mutant, the wurm mutant, is shown as a scanning electronmicroscope picture. d | Inside a cell of one of the distorted mutants, the crooked mutant, bothgrowing regions (left panel) and non-growing regions (right panel) show actin and Golgi vesicles.The difference is that growing regions have fine actin, whereas non-growing regions have bundledactin. This indicates that fusion of vesicles with the membrane is only promoted in regions withfine actin. Online supplementary information S2 (movie) and S3 (movie) show actin-basedmovement in peroxisomes, which, despite the actin-organization defects, is not generally affectedin the mutants (see online supplementary information S3 (movie)) compared with the wild-typecells (see online supplementary information S2 (movie)). The images are reproduced withpermission as follows: parts a and b from REF. 79 © (2003) The American Society of PlantBiologists; part c from REF. 76 © (2003) Springer–Verlag GmbH; the image in part d from REF. 78 ©(2003) The Company of Biologists Ltd.

a b c

d

© 2004 Nature Publishing Group

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 5 | JUNE 2004 | 479

R E V I E W S

well as still unknown processes such as those controlledby STI. Each group of genes has opened new researchareas in the plant sciences and it will be interesting to seewhether the common branching phenotype will tie theseprocesses together. With the discovery that cell-expan-sion genes encode components of the ARP2/3 complex,key components that regulate actin-based growth havebeen identified and will allow the study of the up- anddownstream regulatory processes in plants. Furtheranalysis of trichomes as a single-cell model system offersthe chance to connect the above-mentioned, seeminglyunrelated, processes in the future.

Therefore, it will be challenging to show not onlythat the inhibitor proteins can move, but also howthis is relevant for patterning.

Several pathways seem to have a role in how theswitch from mitosis to endoreduplication and the cyclenumber are controlled. The isolation of further genes, incombination with trichome-specific overexpressionapproaches, should be a valuable addition to the cell-cycle field. The analysis of branching genes has led to theidentification of proteins that are involved in processes asdifferent as intracellular transport, cell-size control, tran-scriptional control and Golgi-dependent processes, as

1. Esau, K. Anatomy of Seed Plants (John Wiley & Sons, New York, 1977).

2. Uphof, J. C. T. Plant hairs (eds. Zimmermann, W. & Ozenda, P. G.) (Gebr. Bornträger, Berlin, 1962).

3. Johnson, H. B. Plant pubescence: an ecologicalperspective. Bot. Rev. 41, 233–258 (1975).

4. Mauricio, R. & Rausher, M. D. Experimental manipulation ofputative selective agents provides evidence for the role ofnatural enemies in the evolution of plant defense. Evolution51, 1435–1444 (1997).

5. Marks, M. D. Molecular genetic analysis of trichomedevelopment in Arabidopsis. Annu. Rev. Plant Physiol. Plant Mol. Biol. 48, 137–163 (1997).

6. Hülskamp, M., Schnittger, A. & Folkers, U. Pattern formationand cell differentiation: trichomes in Arabidopsis as a geneticmodel system. Int. Rev. Cytol. 186, 147–178 (1999).

7. Hülskamp, M., Misera, S. & Jürgens, G. Genetic dissectionof trichome cell development in Arabidopsis. Cell 76,555–566 (1994).Trichome development was dissected into discretedevelopmental steps using a systematic mutagenesisscreen.

8. Larkin, J. C., Young, N., Prigge, M. & Marks, M. D. Thecontrol of trichome spacing and number in Arabidopsis.Development 122, 997–1005 (1996).Elegant demonstration that trichome spacing doesnot involve cell lineage.

9. Melaragno, J. E., Mehrotra, B. & Coleman, A. W.Relationship between endopolyploidy and cell size inepidermal tissue of Arabidopsis. Plant Cell 5, 1661–1668(1993).

10. Folkers, U., Berger, J. & Hülskamp, M. Cell morphogenesisof trichomes in Arabidopsis: differential control of primaryand secondary branching by branch initiation regulators and cell growth. Development 124, 3779–3786 (1997).

11. Schnittger, A., Folkers, U., Schwab, B., Jürgens, G. &Hülskamp, M. Generation of a spacing pattern: the role ofTRIPTYCHON in trichome patterning in Arabidopsis. PlantCell 11, 1105–1116 (1999).

12. Hülskamp, M. & Schnittger, A. Spatial regulation of trichomeformation in Arabidopsis thaliana. Semin. Cell Dev. Biol. 9,213–220 (1998).

13. Scheres, B. Plant patterning: TRY to inhibit your neighbors.Curr. Biol. 12, R804–R806 (2002).

14. Schiefelbein, J. Cell-fate specification in the epidermis: acommon patterning mechanism in the root and shoot. Curr. Opin. Plant Biol. 6, 74–78 (2003).

15. Larkin, J. C., Brown, M. L. & Schiefelbein, J. How do cellsknow what they want to be when they grow up? Lessonsfrom epidermal patterning in Arabidopsis. Annu. Rev. Plant Biol. 54, 403–430 (2003).

16. Koornneef, M., Dellaert, L. W. M. & van der Veen, J. H.EMS- and radiation-induced mutation frequencies atindividual loci in Arabidopsis thaliana. Mutat. Res. 93,109–123 (1982).

17. Koornneef, M. The complex syndrome of ttg mutants.Arabidopsis Information Service 18, 45–51 (1981).

18. Zhang, F., Gonzalez, A., Zhao, M., Payne, C. T. & Lloyd, A. A network of redundant bHLH proteins functions in allTTG1-dependent pathways of Arabidopsis. Development130, 4859–4869 (2003).Excellent analysis of how the many functions of TTG1are mediated by a partially redundant network ofbHLH and MYB transcription factors.

19. Schellmann, S. et al. TRIPTYCHON and CAPRICE mediatelateral inhibition during trichome and root hair patterning inArabidopsis. EMBO J. 21, 5036–5046 (2002).

20. Wada, T., Tachibana, T., Shimura, Y. & Okada, K. Epidermal cell differentiation in Arabidopsis determined by a myb homolog, CPC. Science 277, 1113–1116 (1997).

21. Kirik, V., Simon, M., Hülskamp, M. & Schiefelbein, J. TheENHANCER OF TRY AND CPC1 (ETC1) gene actsredundantly with TRIPTYCHON and CAPRICE in trichomeand root hair cell patterning in Arabidopsis. Dev. Biol. 268,506–513 (2004).

22. Payne, C. T., Zhang, F. & Lloyd, A. M. GL3 encodes a bHLHprotein that regulates trichome development in Arabidopsisthrough interaction with GL1 and TTG1. Genetics 156,1349–1362 (2000).

23. Oppenheimer, D. G., Herman, P. L., Sivakumaran, S., Esch, J.& Marks, M. D. A myb gene required for leaf trichomedifferentiation in Arabidopsis is expressed in stipules. Cell 67, 483–493 (1991).

24. Walker, A. R. et al. The TRANSPARENT TESTA GLABRA1locus, which regulates trichome differentiation andanthocyanin biosynthesis in Arabidopsis, encodes a WD40repeat protein. Plant Cell 11, 1337–1349 (1999).

25. Szymanski, D. B., Lloyd, A. M. & Marks, M. D. Progress inthe molecular genetic analysis of trichome intiation andmorphogenesis in Arabidopsis. Trends Plant Sci. 5,214–219 (2000).

26. Esch, J. J. et al. A contradictory GLABRA3 allele helpsdefine gene interactions controlling trichome development inArabidopsis. Development 130, 5885–5894 (2003).First insight into how the negative regulator TRYmight counteract the GL1–GL3–TTG1 complex. Yeastthree-hybrid analysis showed that TRY competes withGL1 for binding to GL3.

27. Wada, T. et al. Role of a positive regulator of root hairdevelopment, CAPRICE, in Arabidopsis root epidermal celldifferentiation. Development 129, 5409–5419 (2002).

28. Lee, M. M. & Schiefelbein, J. Cell patterning in theArabidopsis root epidermis determined by lateral inhibitionwith feedback. Plant Cell 14, 611–618 (2002).Elegant molecular-genetic study that is the first toshow a regulatory feedback mechanism during root-hair patterning.

29. Rerie, W. G., Feldmann, K. A. & Marks, M. D. The glabra 2gene encodes a homeo domain protein required for normaltrichome development in Arabidopsis. Genes Dev. 8,1388–1399 (1994).

30. Cristina, M. D. et al. The Arabidopsis Athb-10 (GLABRA2) isan HD-Zip protein required for regulation of root hairdevelopment. Plant J. 10, 393–402 (1996).

31. Szymanski, D. B., Jilk, R. A., Pollock, S. M. & Marks, M. D.Control of GL2 expression in Arabidopsis leaves andtrichomes. Development 125, 1161–1171 (1998).

32. Ohashi, Y., Ruberti, I., Morelli, G. & Aoyama, T. Entopicallyadditive expression of GLABRA2 alters the frequency andspacing of trichome initiation. Plant J. 21, 5036–5046(2002).

33. Ohashi, Y. et al. Modulation of phospholipid signaling byGLABRA2 in root-hair pattern formation. Science 300,1427–1430 (2003).Identification of the first gene that functionsdownstream of the patterning machinery.

34. Walker, J. D., Oppenheimer, D. G., Concienne, J. & Larkin, J. C. SIAMESE, a gene controlling theendoreduplication cell cycle in Arabidopsis thalianatrichomes. Development 127, 3931–3940 (2000).

35. Schnittger, A., Schobinger, U., Stierhof, Y. D. & Hülskamp, M.Ectopic B-type cyclin expression induces mitotic cycles inendoreduplicating Arabidopsis trichomes. Curr. Biol. 12,415–420 (2002).

36. Schnittger, A. et al. Ectopic D-type cyclin expressioninduced not only DNA replication but also cell division inArabidopsis trichomes. Proc. Natl Acad. Sci. USA 99,6410–5415 (2002).Overexpression studies showed that plant D-typecyclins function, not only at the G1 transition, but alsoat the entry of mitosis.

37. Sugimoto-Shirasu, K., Stacey, N. J., Corsar, J., Roberts, K.& McCann, M. C. DNA topoisomerase VI is essential forendoreduplication in Arabidopsis. Curr. Biol. 12, 1782–1786(2002).

38. Jacobsen, S. E., Binkowski, K. A. & Olszewski, N. E.SPINDLY, a tetratricopeptide repeat protein involved ingibberellin signal transduction in Arabidopsis. Proc. NatlAcad. Sci. USA 93, 9292–9296 (1996).

39. Chien, J. C. & Sussex, I. M. Differential regulation oftrichome formation on the adaxial and abaxial leaf surfacesby gibberellins and photoperiod in Arabidopsis thaliana (L.)Heynh. Plant Physiol. 111, 1321–1328 (1996).

40. Telfer, A., Bollman, K. M. & Poethig, R. S. Phase change andthe regulation of trichome distribution in Arabidopsisthaliana. Development 124, 645–654 (1997).

41. Downes, B. P., Stupar, R. M., Gingerich, D. J. & Vierstra, R. D.The HECT ubiquitin-protein ligase (UPL) family inArabidopsis: UPL3 has a specific role in trichomedevelopment. Plant J. 35, 729–742 (2003).

42. El Refy, A. et al. The Arabidopsis KAKTUS gene encodes aHECT protein and controls the number of endoreduplicationcycles. Mol. Genet. Genomics 270, 403–414 (2004).

43. Hülskamp, M. How plants split hairs. Curr. Biol. 10,R308–R310 (2000).

44. Krishnakumar, S. & Oppenheimer, D. G. Extragenicsuppressors of the Arabidopsis zwi-3 mutation identify new genes that function in trichome branch formation andpollen tube growth. Development 126, 3079–3088 (1999).

45. Oppenheimer, D. Genetics of plant cell shape. Curr. Opin.Plant Biol. 1, 520–524 (1998).

46. Luo, D. & Oppenheimer, D. G. Genetic control of trichomebranch number in Arabidopsis: the roles of the FURCA loci.Development 126, 5547–5557 (1999).Detailed genetic studies on the genetic interactionsbetween the branching genes showed that theyfunction largely in independent pathways.

47. Perazza, D. et al. Trichome cell growth in Arabidopsisthaliana can be depressed by mutations in at least fivegenes. Genetics 152, 461–476 (1999).

48. Kirik, V. et al. CPR5 is involved in cell proliferation and celldeath control and encodes a novel transmembrane protein.Curr. Biol. 11, 1891–1895 (2001).

49. Mathur, J. & Chua, N.-H. Microtubule stabilization leads togrowth reorientation in Arabidopsis thaliana trichomes. Plant Cell 12, 465–477 (2000).

50. Kirik, V. et al. The Arabidopsis TUBULIN-FOLDINGCOFACTOR A gene is involved in the control of the α/β-tubulin monomer balance. Plant Cell 14, 2265–2276(2002).

51. Kirik, V. et al. Functional analysis of the tubulin-foldingcofactor C in Arabidopsis thaliana. Curr. Biol. 12,1519–1523 (2002).

52. Steinborn, K. et al. The Arabidopsis PILZ group genesencode tubulin-folding cofactor orthologs required for celldivision but not cell growth. Genes Dev. 16, 959–971 (2002).

53. Bichet, A., Desnos, T., Turner, S., Grandjean, O. & Höfte, H.BOTERO1 is required for normal orientation of corticalmicrotubules and anisotropic cell expansion in Arabidopsis.Plant J. 25, 137–148 (2001).

© 2004 Nature Publishing Group

480 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/molcellbio

R E V I E W S

54. Webb, M., Jouannic, S., Foreman, J., Linstead, P. & Dolan, L. Cell specification in the Arabidopsis root epidermisrequires the activity of ECTOPIC ROOT HAIR 3 — a katanin-p60 protein. Development 129, 123–131 (2002).

55. Burk, D. H., Liu, B., Zhong, R., Morrison, W. H. & Ye, Z. H. A katanin-like protein regulates normal cell wall biosynthesisand cell elongation. Plant Cell 13, 807–827 (2001).

56. Torres-Ruiz, R. A. & Jürgens, G. Mutations in the FASS geneuncouple pattern formation and morphogenesis inArabidopsis development. Development 120, 2967–2978(1994).

57. Traas, J. et al. Normal differentiation patterns in plantslacking microtubular preprophase bands. Nature 375,676–677 (1995).

58. Camilleri, C. et al. The Arabidopsis TONNEAU2 geneencodes a putative novel protein phosphatase 2A regulatorysubunit essential for the control of the cortical cytoskeleton.Plant Cell 14, 833–845 (2002).

59. Qiu, J. L., Jilk, R., Marks, M. D. & Szymanski, D. B. TheArabidopsis SPIKE1 gene is required for normal cell shapecontrol and tissue development. Plant Cell 14, 101–118(2002).

60. Brugnera, E. et al. Unconventional Rac-GEF activity ismediated through the Dock180–ELMO complex. NatureCell Biol. 4, 574–582 (2002).

61. Deeks, M. J. & Hussey, P. J. Arp2/3 and ‘the shape of thingsto come’. Curr. Opin. Plant Biol. 6, 561–567 (2003).Excellent review on the possible role and regulation ofactin through the ARP2/3 complex.

62. Reddy, A. S., Narasimhulu, S. B., Safadi, F. & Golovkin, M. A plant kinesin heavy chain-like protein is a calmodulin-binding protein. Plant J. 10, 9–21 (1996).

63. Reddy, A. S. N., Safadi, F., Narasimhulu, S. B., Golovkin, M.& Hu, X. A novel plant calmodulin-binding protein with akinesin heavy chain motor domain. J. Biol. Chem. 271,7052–7060 (1996).

64. Reddy, A. S. N., Narasimhulu, S. B. & Day, I. S. Structuralorganization of a gene encoding a novel calmodulin-bindingkinesin-like protein from Arabidopsis. Gene 204, 195–200(1997).

65. Song, H., Golovkin, M., Reddy, A. S. & Endow, S. A. In vitro motility of AtKCBP, a calmodulin-binding kinesinprotein of Arabidopsis. Proc. Natl Acad. Sci. USA 94,322–237 (1997).

66. Deavours, B. E., Reddy, A. S. & Walker, R. A.Ca2+/calmodulin regulation of the Arabidopsis kinesin-likecalmodulin-binding protein. Cell Motil. Cytoskeleton 40,408–416 (1998).

67. Oppenheimer, D. G. et al. Essential role of a kinesin-likeprotein in Arabidopsis trichome morphogenesis. Proc. NatlAcad. Sci. USA 94, 6261–6266 (1997).

68. Reddy, V. S., Day, I., Thomas, T. & Reddy, A. S. N. KIC, anovel Ca2+ binding protein with one EF-hand motif, interactswith a microtubule motor protein and regulates trichomemorphogenesis. Plant Cell 16, 185–200 (2004).

69. Folkers, U. et al. The cell morphogenesis geneANGUSTIFOLIA encodes a CtBP/BARS-like protein and isinvolved in the control of the microtubule cytoskeleton.EMBO J. 21, 1280–1288 (2002).

70. Kim, G. T. et al. The ANGUSTIFOLIA gene of Arabidopsis, aplant CtBP gene, regulates leaf-cell expansion, the

arrangement of cortical microtubules in leaf cells andexpression of a gene involved in cell-wall formation. EMBO J. 26, 1267–1279 (2002).

71. Nibu, Y., Zhang, H. & Levine, M. Interaction of a short-rangerepressors with Drosophila CtBP in the embryo. Science280, 101–104 (1998).

72. Matteis, M. D. et al. Stimulation of endogenous ADP-ribosylation by brefeldin A. Proc. Natl Acad. Sci. USA 91,1114–1118 (1994).

73. Lippincott-Schwartz, J., Yuan, L. C., Bonifacino, J. S. &Klausner, R. D. Rapid redistribution of Golgi proteins into theER in cells treated with brefeldin A: evidence for membranecycling from Golgi to ER. Cell 56, 801–813 (1989).

74. Ilgenfritz, H. et al. The Arabidopsis STICHEL gene is aregulator of trichome branch number and encodes a novelprotein. Plant Physiol. 131, 643–655 (2003).

75. Szymanski, D. B., Marks, M. D. & Wick, S. M. Organized F-actin is essential for normal trichome morphogenesis inArabidopsis. Plant Cell 11, 2331–2348 (1999).

76. Schwab, B. et al. Regulation of cell expansion by theDISTORTED genes in Arabidopsis thaliana: actin controlsthe spatial organization of microtubules. Mol. Genet.Genomics 269, 350–360 (2003).

77. Mathur, J., Spielhofer, P., Kost, B. & Chua, N.-H. The actincytoskeleton is required to elaborate and maintain spatialpatterning during trichome cell morphogenesis inArabidopsis thaliana. Development 126, 5559–5568 (1999).

78. Mathur, J. et al. Arabidopsis CROOKED encodes for thesmallest subunit of the ARP2/3 complex and controls cellshape by region specific fine F-actin formation.Development 130, 3137–3146 (2003).

79. Mathur, J., Mathur, N., Kernebeck, B. & Hülskamp, M.Mutations in actin-related proteins 2 and 3 affect cell shapedevelopment in Arabidopsis. Plant Cell 15, 1632–1645(2003).

80. Le, J., El-Assal Sel, D., Basu, D., Saad, M. E. & Szymanski, D. B. Requirements for Arabidopsis ATARP2and ATARP3 during epidermal development. Curr. Biol. 13,1341–1347 (2003).

81. Li, S., Blanchoin, L., Yang, Z. & Lord, E. M. The putativeArabidopsis arp2/3 complex controls leaf cellmorphogenesis. Plant Physiol. 132, 2034–2044 (2003).

82. Mullins, R. D., Heuser, J. A. & Pollard, T. D. The interaction ofArp2/3 complex with actin: nucleation, high affinity pointedend capping, and formation of branching networks offilaments. Proc. Natl Acad. Sci. USA 95, 6181–6186 (1998).

83. Svitkina, T. M. & Borisy, G. G. ARP2/3 complex and actindepolymerizing factor/cofilin in dendritic organization andtreadmilling of actin filament array in lamellipodia. J. Cell Biol.145, 1009–1026 (1999).

84. Mathur, J. & Hülskamp, M. Signal transduction: Rho-likeproteins in plants. Curr. Biol. 12, R526–R528 (2002).

85. Smith, L. G. Cytoskeletal control of plant cell shape: gettingthe fine points. Curr. Opin. Plant Biol. 6, 63–73 (2003).

86. Yang, Z. Small GTPases: versatile signaling switches inplants. Plant Cell 14 (Suppl.) 375–388 (2002).

87. De Veylder, L. et al. Functional analysis of cyclin-dependentkinase inhibitors of Arabidopsis. Plant Cell 13, 1653–1668(2001).

88. Hiromura, K., Pippin, J. W., Fero, M. L., Roberts, J. M. &Shankland, S. J. Modulation of apoptosis by the cyclin-

dependent kinase inhibitor p27(Kip1). J. Clin. Invest. 103,597–604 (1999).

89. Wang, H. et al. ICK1, a cyclin-dependent protein kinaseinhibitor from Arabidopsis thaliana interacts with both Cdc2aand CycD3, and its expression is induced by abscisic acid.Plant J. 15, 501–510 (1998).

90. Jasinski, S. et al. The CDK inhibitor NtKIS1a is involved inplant development, endoreduplication and restores normaldevelopment of cyclin D3;1-overexpressing plants. J. Cell Sci. 115, 973–982 (2002).

91. Schnittger, A., Weinl, C., Bouyer, D., Schobinger, U. &Hülskamp, M. Misexpression of the cyclin-dependent kinaseinhibitor ICK1/KRP1 in single-celled Arabidopsis trichomesreduces endoreduplication and cell size and induces celldeath. Plant Cell 15, 303–315 (2003).

92. Glazebrook, J. Genes controlling expression of defenseresponses in Arabidopsis — 2001 status. Curr. Opin. Plant Biol. 4, 301–308 (2001).

93. Haughn, G. W. & Somerville, C. R. Genetic control ofmorphogenesis in Arabidopsis. Dev. Genet. 9, 73–89(1988).

94. Potikha, T. & Delmer, D. A mutant of Arabidopsis thalianadisplaying altered patterns of cellulose deposition. Plant J. 7, 453–460 (1995).

95. Meinhardt, H. & Gierer, A. Applications of a theory ofbiological pattern formation based on lateral inhibition. J. Cell Sci. 15, 321–346 (1974).

96. Vos, J. W., Safadi, F., Reddy, A. S. & Hepler, P. K. Thekinesin-like calmodulin binding protein is differentiallyinvolved in cell division. Plant Cell 12, 979–990 (2000).

97. Bowser, J. & Reddy, A. S. N. Localization of a kinesin-likecalmodulin-binding protein in dividing cells of Arabidopsisand tobacco. Plant J. 12, 1429–1437 (1997).

98. Schwab, B., Folkers, U., Ilgenfritz, H., and Hülskamp, M.Trichome morphogenesis in Arabidopsis. Philos. Trans. R.Soc. Lond. B 355, 879–883 (2000).

AcknowledgementsI would like to thank H. Meinhardt for providing the images and themovie that are presented in BOX 1 and for stimulating discussions. Iwould also like to thank the members of the laboratory for helpfulcomments on the manuscript. Research in the author’s laboratoryis supported by the Deutsche Forschungsgemeinschaft and theVolkswagen Stiftung.

Competing interests statementThe author declares that he has no competing financial interests.

Online links

DATABASESThe following terms in this article are linked online to:TAIR: http://www.arabidopsis.org/AN | CPC | CPR5 | CYCB1;2 | CYCD3;1 | EGL3 | GL1 | GL2 | GL3 |HYP6 | KAK | RHL2 | SIM | STI | TTG1 | WER | ZWI

SUPPLEMENTARY INFORMATIONSee online article: S1 (movie) | S2 (movie) | S3 (movie)Access to this links box is available online.

© 2004 Nature Publishing Group