_____________________________ _____________________________
Functional Characterization of the Pointed Cotyledon Subclass
of HDZip Genes in Arabidopsis thaliana
BY
JOHANNES HANSON
ACTA UNIVERSITATIS UPSALIENSIS UPPSALA 2000
ABSTRACT
Hanson, J. 2000. Functional Characterization of the Pointed Cotyledon Subclass of HDZip genes in Arabidopsis thaliana. Acta Universitatis Upsaliensis. Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 580.
56 pp. Uppsala. ISBN 91-554-4846-1.
Genes encoding homeodomain leucine zipper, HDZip, transcription factors constitute a large gene family in Arabidopsis thaliana. In this thesis the isolation and characterization of four HDZip genes (ATHB3, -13, -20 and –23) is described. These genes are similar in sequence and form a distinct subclass within the HDZip gene family. Since the genes cause similar alterations in cotyledon shape when expressed constitutively, we refer to the members of this subclass as the pointed cotyledon HDZip genes.
To determine the biological functions of the genes, the phenotypes of plants constitutively expressing the genes have been analysed. Each of the genes specifically inhibits lateral cell expansion in cotyledons and leaves, and thereby causes them to be abnormally narrow. Detailed expression analysis shows that only ATHB23 is expressed in the entire leaf and cotyledon from early stages of development while ATHB20 is predominantly expressed in the root cortex. ATHB13 is expressed in basal parts of mature leaves and floral organs and ATHB3 in root and stem cortex. The ATHB13 protein acts within a signalling pathway that mediates a response to sucrose that specifically regulates the expression of specific sugar-regulated genes. ATHB3 specifically inhibits primary root development without affecting the development of secondary roots when constitutively expressed.
Reduced expression of ATHB3 by antisense suppression results in increased expression of ATHB13, indicating that ATHB3 acts as a repressor of ATHB13 expression in the wild type.
This thesis also reports the isolation of seven new genes of HDZip class I and reviews available functional information on the genes in this class. One conclusion is that HDZip I proteins that are closely related phylogenetically are also functionally related, in most cases. Seven different mutations in HDZip I genes were identified.
The lack of phenotypic deviations from wild type of these mutants suggests that these HDZip proteins act in a redundant fashion in the plant.
Johannes Hanson, Department of Physiological Botany, Evolutionary Biology Centre, Villav. 6, SE-752 36, Uppsala, Sweden
© Johannes Hanson 2000 ISSN 1104-232X
ISBN 91-554-4846-1
Printed in Sweden by University Printers, Uppsala 2000
I Hanson, J., Johannesson, H., and Engström, P. (2000). Sugar dependent alterations in cotyledon and leaf development in transgenic plants expressing the HDZip gene ATHB13. Plant Mol. Biol. In press
II Hanson, J., Regan, S., and Engström, P. (2000). ATHB13 is highly expressed in the vascular tissue at the base of petioles in both Arabidopsis and hybrid aspen.
(manuscript)
III Hanson, J., and Engström, P. (2000). Constitutive expression of each of four closely related homeobox genes in transgenic Arabidopsis causes similar pointed cotyledon phenotypes. (manuscript)
IV Hanson, J., and Engström, P. (2000). ATHB3 represses ATHB13 expression, and, when constitutively expressed, specifically affects primary root development in Arabidopsis thaliana. (manuscript)
V Johannesson, H., Hanson, J., Söderman, E., Wang, Y., and Engström, P. (2000) HDZip proteins in Arabidopsis thaliana: a case of functional conservation and redundancy within a family of transcription factors. (manuscript)
Manuscript I has been reprinted with the kind permission of Kluwer Academic Publishers.
ABBREVIATIONS... 6
PREFACE ... 7
INTRODUCTION ... 8
Arabidopsis – a useful weed ... 8
Root Development ... 9
Post-embryonic development of the root... 9
The pattern for post-embryonic development of the root is laid down during embryogenesis... 10
Initiation and formation of secondary roots ... 12
Cotyledon and leaf development... 13
Life history of a leaf ... 13
The control of leaf expansion and leaf shape ... 13
The cotyledon ... 14
Sugar sensing ... 16
HDZip transcription factors ... 19
The pre-history of the HDZip domain... 19
HDZip transcription factors are encoded by a large gene family in Arabidopsis... 20
HDZip proteins act as dimeric transcription factors... 21
HDZip proteins are involved in a wide range of processes in plants ... 24
RESULTS AND DISCUSSION Isolation and characterization of novel HDZip genes... 26
Novel genes distantly related to previously known HDZip I genes ... 26
Pointed cotyledon-HDZip genes ... 28
Functional characterization of the poc-HDZip genes ... 30
Constitutive expression of poc-HDZip genes results in pointed cotyledons and serrated leaves ... 30
Poc-HDZip genes are differentially expressed... 31
ATHB13 affects cotyledon and leaf development in a sucrose dependent manner... 33
ATHB3 specifically affects primary root development when constitutively expressed... 35
ATHB3 antisense gene expression increases ATHB13 expression to higher levels ... 36
Lack of phenotypic deviations caused by HDZip I gene mutations indicates functional redundancy within the gene family... 37
Concluding remarks... 38
SUMMARY IN SWEDISH – POPULÄRVETENSKAPLIG SAMMANFATTNING ... 40
ACKNOWLEDGEMENTS ... 42
REFERENCES ... 44
2dGlc 2-deoxy glucose
35S::ATHB3 line constitutively expressing the gene ATHB3
35S::POC line constitutively expressing one of the poc-HDZip genes 3-O-mGlc 3-O-methyl glucose
6dGlc 6-deoxy glucose
ABA abscisic acid
ACC amino-cyclopropane-carboxylic acid (ethylene precursor) AGI Arabidopsis Genome Initiative
ATHB Arabidopsis thaliana homeobox ATP adenosine triphosphate
BAP 6-benzylaminopurine (cytokinin)
bp base pair
bZip basic leucine zipper
cDNA complementary DNA
DNA deoxyribonucleic acid HDZip homeodomain leucine zipper
HXK hexokinase
HXT hexose transporter
IAA 3-indoleacetic acid (auxin)
mRNA messenger RNA
PCR polymerase chain reaction poc-HDZip pointed cotyledon HDZip RING really interesting new gene RNA ribonucleic acid
SAM shoot apical meristem SUT sucrose transporter T-DNA transferred DNA
The following conventions have been followed in this thesis:
Names of genes are written in italicized upper-case letters, e.g. ATHB13
Names of proteins are written in non-italicized upper-case letters, e.g. ATHB13 Names of mutants and mutations are written in italicized lower-case letters, e.g. athb13-1
In 1993, I joined the HDZip group of Peter Engström at the department of physiological botany, Uppsala University. At that time, only four HDZip genes had been cloned in the lab and I thought we were going to show that these homeobox genes regulated similar processes to their animal counterparts. I now know that this was a naive belief of a young developmental biologist who was unaware of how plants live and develop.
Over the years here in Uppsala I have gradually gained more and more understanding of plant physiology and development. In the beginning I was very disappointed and desperately tried to fit plants into my preconceptions of development, based on what I had learned about the development of the fly Drosophila melanogaster and other animal models. I have now realized that the cute little plant Arabidopsis thaliana is not just a simplified fruit fly, but rather an elegant survivor that has evolved its own fascinating systems to cope with an ever-changing environment. However, this thesis does not reflect my journey in plant biology. It is just a snapshot of my present location. I would like to dedicate this snapshot to the species Heliathus tuberosus.
Uppsala, 2000 Johannes Hanson
Arabidopsis – a useful weed
Arabidopsis thaliana (L.) Heynh, commonly known as thale cress, mouse-ear cress or wall cress, but usually referred to as Arabidopsis in the scientic literature, is a plain plant that only attracts the eyes of researchers (Figure 1). It belongs to the mustard family (Brassicaceae or Cruciferae) and is related to various economically important crops such as rape and broccoli. It is widely distributed in the temperate climate zone of the Northern Hemisphere and is usually found in poor and exposed habitats such as roadsides (Price et al. 1994). In nature, it grows as a winter annual.
The seed germinates and grows during the autumn, the plant survives the winter as a rosette, then owers and sets seeds in the spring (Rédei 1992).
Arabidopsis has become the favourite model organism for plant research. It was
rst recognized as an organism suited for genetic investigations in the middle of the twentieth century (the early history of Arabidopsis as a scientic model is reviewed by Rédei 1992). Geneticists found it convenient for many reasons: it has a short generation time of about six weeks, it is small in size (its rosette diameter being approximately 5 cm and its inorescence height about 30 cm) and can be grown in large quantities, it produces large numbers of seeds (up to 5000 per plant), it is self pollinating but can be cross-pollinated with ease and it is easy to mutagenize by either chemicals or radiation (Koncz et al. 1992). When the breakthrough of molecular genetics came in the 1980’s there was at rst no consensus on which organism to work on as a model. Investigations were focused on many different organisms including tomato, petunia, pea, rice, barley, maize, snapdragon and tobacco - all of which have different advantages and disadvantages. Although an impressive amount of information was collected, advances in many disciplines were limited because information was scattered, comparison of results from different organisms was often difcult and efforts were sometimes duplicated as the same information was collected from many organisms (Meinke et al. 1998). A search for a new organism suited to the new molecular and classical genetical methods was initiated and, through a gradual process, Arabidopsis was chosen. The new generation of molecular geneticists favoured Arabidopsis for the same reasons as the previous generation. The newcomers also beneted from the work that had already been done on the organism, like the genetic maps and hundreds of characterized mutants (Meinke et al. 1998; Somerville and Meyerowitz 1994). Arabidopsis was also found to be easily transformed by means of Agrobacterium tumefaciens (Bechtold et al. 1993). Furthermore, it has a very small genome of only approximately 120 million base pairs (bp) and a low level of repetitive DNA (Meyerowitz 1992). The Arabidopsis genome contains a gene every 5,000 bp on average and is estimated to contain approximately 20,000 genes (Bevan et al. 1998). In 1996, the Arabidopsis Genome Initiative, AGI, was established to
facilitate the sequencing of the whole Arabidopsis genome (Bevan et al. 1997), a task that will be completed before the end of the year 2000. Thousands of mutants have been generated by means of insertion mutagenesis. These mutants are easily screened by means of PCR so, within a few years, knockout mutants for virtually every Arabidopsis gene will be available to the scientic community (Krysan et al.
1999).
The post-genome era of Arabidopsis research will surely be as successful as the rst fty years and certainly more and more of the information collected from Arabidopsis will nd uses in new improvements of more economically important species.
Root Development
Post-embryonic development of the root
Arabidopsis has been shown to be an excellent system for studying post-embryonic development of the root, as it has a fairly uncomplicated cellular organization (Dolan et al. 1993) in which the developmental history of every cell has been determined (Scheres et al. 1994), and its roots are nearly transparent. The Arabidopsis root consists of a few different cell types organized in a radially symmetrical pattern along the length axis. The cell types of each layer are continuously produced in les from progenitor cells, the initials, similar to a conveyor belt, as the root elongates (Figure 2). Only four cells of the Arabidopsis root meristem, the quiescent centre, are fully undifferentiated and these cells rarely divide under normal conditions. The cells around the quiescent centre are the initials for the different cell-types of the
Figure 1
Arabidopsis thaliana var. Ler . Drawing by Dr. Eva Sundberg.
root (the epidermis, cortex, endodermis, etc., see Figure 2). The initials divide to form one cell that takes on the fate the mother cell and a second cell that is destined to differentiate into a certain type (epidermis, cortex, endodermis, etc.). Mutations affecting these processes in different cell types have been identied, for example the shortroot mutation that blocks the formation of endodermis cells (Scheres et al.
1995). The fate of each initial cell is determined by its position in the root: positional signals pass from differentiated cells to undifferentiated cells rather than from the initials (van den Berg et al. 1995). Thus, cells differentiate like their more basal daughters. The molecular nature of these positional cues is not known but mapping the plasmodesmata in the root has shown that connections are preferentially made to cells of the same type in the root (van den Berg et al. 1998 and references therein).
The contacts between the initials and the quiescent centre cells keep the initials in an undifferentiated state since ablation of the cells in the quiescent centre causes initials to stop dividing and to differentiate according to cues from the daughter cells in the same le (van den Berg et al. 1997).
The pattern for post-embryonic development of the root is laid down during embryogenesis
The cellular organization of the root is laid down during embryogenesis, and the ontogeny of each cell has been determined through the use of genetically marked embryonic cells, Figure 2 (Scheres et al. 1994). The rst division of the zygote generates an apical and a basal cell and the root is formed from derivatives of both of these cells. The basal cell generates the quiescent centre and the columella root cap initials. The rest of the root cells, including the initials, are derivatives of the apical cell (reviewed in Scheres and Heidstra 1999). The apical-basal pattern is laid down early in embryogenesis as dened by the rst visible developmental deviations detected in the monopteros, mp, mutant (Berleth and Jurgens 1993). The capacity of mp mutant plants to make adventitious roots indicates that mp plants can make largely normal apical structures. All organs, however, display defective vascular strands and impaired auxin transport (Przemeck et al. 1996). This indicates that MP promotes axialization and cell le formation, processes that are important for both embryonic axis formation and vascular system development, rather than specifying the apical-basal pattern. The determination of cell fates along the apical-basal axis also involves the HOBBIT, HBT gene, as the hbt mutant is unable to form a root meristem (Willemsen et al. 1998). From very early development the cells that will give rise to the quiescent centre and the columella initials (Figure 2) divide aberrantly in the hbt mutant, consistent with the hypothesis that the gene has a specic function
-
A
B
C
D
E
stele
endod.
cortex epid.
l.r.c.
col.
cot. cot.
Figure 2
Embryonic and post-embryonic root development of Arabidopsis
Simplified schematic drawings of cellular arrangements (transverse sections) during root
development. Cortex and endodermis cells are depicted in dark grey and the stele (internal), columella and lateral root cap (peripheral) is shown in light grey. The quiescent centre and epidermis of the root are in white, as are all other cells of the non-root lineage. The figures are adopted from Scheres et al.
1994 and van den Berg et al. 1998, with kind permission from the authors.
A, Early heart stage embryo. B, Late heart stage embryo. C, Seedling. cot., cotyledon. D, Root apex of seedling. E, Close up of central region in root meristem shown in D. Initial cells for all the different cell types surrounding the quiescent centre. endod., endodermis; epid., epidermis; l.r.c., lateral root cap; col., columella. Arrows indicate the direction of daughter cell displacement.
in these cells (Willemsen et al. 1998). The radial axis of the root is laid down after the apical-basal axis, since the rst deviations are detectable in mutants specically affected in stages of radial pattern formation that appear later in embryogenesis than those altered in hbt and mp (Scheres et al. 1995). It is reasonable to believe that embryonic root formation shares extensive similarity to post-embryonic root development as hbt and mutations affecting the radial pattern of the root also cause the same deviations in secondary roots (Malamy and Benfey 1997a).
Initiation and formation of secondary roots
The primary root of the adult Arabidopsis plant is only a minor part of its root system, which is largely composed of secondary roots formed from the primary root or other tissue, and further secondary roots formed from them, and so on. The secondary roots perform and develop like primary roots, but are not initiated during embryogenesis.
Secondary roots are initiated from differentiated tissue and the frequency of initiation is highly inuenced by environmental conditions (Charlton 1996). This plasticity in the formation of the root system is one way in which plants overcome their inability to move away from poor soils and towards nutrients. The lateral roots of Arabidopsis are initiated from cells in the pericycle (an internal layer adjacent to the stele, Figure 2) of the root. A small number of pericycle cells start dividing and eventually form the lateral root primordium. The lateral root primordium forms a structure identical to the root meristem through a series of dened stages of development, and emerges through the outer cell layers of the root (Laskowski et al. 1995; Malamy and Benfey 1997b). Most mutations identied in Arabidopsis that affect the development of the primary root also affect the lateral roots. This indicates that the formation of the lateral root meristem shares many regulatory mechanisms with those of primary root initiation during embryogenesis (reviewed by Malamy and Benfey 1997a; Scheres and Heidstra 1999).
The induction of lateral roots is dependent on the hormone auxin, which is transported from aerial parts of the plant. This has been demonstrated by the application of both auxin (Torrey 1950) and inhibitors of auxin transport (Reed et al. 1998). Mutations with reduced sensitivity to auxin exhibit a reduction or loss of lateral roots (Celenza et al. 1995; Estelle and Sommerville 1987) and the extensive production of lateral roots in the mutant alf1 (Celenza et al. 1995), which is allelic to superroot and rooty, has been linked to elevated auxin levels (Boerjan et al.
1995; King et al. 1995). In screens for mutants with altered frequencies of lateral root initiation in Arabidopsis, only one mutation, alf4-1, has been shown to inhibit this induction without affecting auxin perception or synthesis (Celenza et al. 1995), demonstrating the importance of the hormone in the process. However, the initiation
of secondary roots in other species, originating from other tissues, may be regulated by other hormones, as illustrated by the ethylene dependence of adventitious root formation (i.e. secondary root production from non-root tissue) from stem cuttings of tomato and petunia (Clark et al. 1999).
Cotyledon and leaf development
Life history of a leaf
All the leaves of a plant originate from a small group of cells in the apical parts of the shoot, the shoot apical meristem (SAM). This group of cells forms a dome that undergoes extensive cell division in a highly regular manner (Lyndon 1970). As new cells are formed the older ones are displaced to the periphery of the dome. These cells will later form the primordia of leaves (Lyndon and Cunninghame 1986). The shoot apical meristem continuously produces leaf primordia at its anks in a spiral pattern. The earliest histological evidence of primordium initiation is a change in the orientation of cell divisions in the region where the primordia will form (Lyndon 1970). At the molecular level this is correlated with a repression of knox genes in maize and Arabidopsis. These genes are expressed in the central parts of the SAM and have been shown to be important for maintaining the undifferentiated state of these cells (Reiser et al. 2000).
During its initiation or early development the leaf primordium is divided into discrete domains along the basic axes of the future leaf, in other words, a polarity is established. The cells in the different positions along these axes/polarities then develop according to different fates. The phantastica mutation in Antirrhinum majus abolishes the formation of both the lateral and the dorsoventral axis as the leaves of the mutant plants have complete radial symmetry (Waites and Hudson 1995).
Regulation of pattern along the proximodistal axis has been extensively studied in maize where the distal and proximal parts of the leaf differ in many aspects of their cellular differentiation (Sylvester et al. 1990). At least 15 different gene products have been shown to specically affect structures along this axis in maize (reviewed in Sylvester et al. 1996).
The control of leaf expansion and leaf shape
Much of the diversity of leaf shapes seen in nature is caused by variation in the amount of expansion within the leaf (Dale 1988). Variations in leaf expansion causing different shapes of leaves may occur early, as shown in snapdragon, Antirrhinum majus, where variation in leaf shape along the length of the shoot is due to variation in the growth of the lamina early in development (Harte and Meinhard 1979a, b). In
Arabidopsis, which has leaves with a shape similar to those of snapdragon, the genes ANGUSTIFOLIA and ROTUNDIFOLIA act after leaf blade formation to control the length and width of the leaf (Tsuge et al. 1996). The leaves and leaf cells of the angustifolia (an) mutant are signicantly narrower than wt leaves (Rédei 1962) but the leaf length is not affected by the mutation (Tsuge et al. 1996). Cell shapes in plants are determined by the direction-specic inhibition of expansion exerted by cellulose microbrils, while the driving force is the positive turgor pressure of the cell (Taiz and Zeiger 1991). This means that inhibition of expansion in one direction by cell wall microbrils leads to expansion in the other directions. AN affects cell width, and secondarily leaf width, by this mechanism as an leaves and cells are signicantly thicker than wt cells (Tsuge et al. 1996). AN also affects the orientation of microtubules and, thus, most likely the orientation of cellulose microbrils (Tsukaya et al. 1999). This mechanism also seems to occur in species other than Arabidopsis since the fat mutant from tobacco, Nicotiana sylvestris, also has thick, narrow leaves (McHale 1993).
The leaves and cells of the rotundifolia3 (rot3) mutant are shorter than those of the wt, but the width and thickness of the cells are not altered (Tsuge et al. 1996). This proves that plants can inhibit leaf cell expansion in one direction without affecting expansion in the other directions. ROT3 encodes a cytochrome P450 with similarities to animal steroid hydroxylases (Kim et al. 1998). How it affects the length of the leaf is not currently known, but the mechanism is distinct from that of AN as the arrangement of microtubili is not altered in the rot3 mutant (Tsukaya et al. 1999).
Over-expression of ROT3 in transgenic plants causes the development of longer leaves compared to wt, indicating that the gene can both positively and negatively regulate leaf length (Kim et al. 1999). Over-expression of the rot3-2 gene (with encodes a ROT3 protein with an amino acid substitution) causes the rot3 phenotype (short leaves) to appear in wt plants, indicating that the ROT3-2 protein dominantly inhibits the wt protein (Kim et al. 1999).
The cotyledon
Unlike the leaf, the cotyledon does not originate from a meristem. The cotyledons are initiated during embryogenesis (Figure 2). The Arabidopsis cotyledons contain stored reserves (lipids, proteins and starch) that are broken down and mobilized during germination and early phases of seedling establishment (Manseld and Briarty 1996). Approximately 60 hours post-imbibition the Arabidopsis seedling switches to auxotrophic growth and starts to photosynthesise. After this transition the cotyledons
physiologically act as leaves (Manseld and Briarty 1996). When the rosette leaves are fully developed, the small cotyledons are effectively shaded from the light and senesce.
The Arabidopsis mutants extra cotyledon1 and -2 (xtc1 and –2) and altered meristem programming 1, amp1, frequently have more than two cotyledons (Conway and Poethig 1997). The position and timing of emergence of these extra cotyledons indicate that the rst leaves of these mutants develop according to cotyledon fate.
However, these cotyledon-like leaves are initiated prematurely during embryogenesis by aberrant timing of SAM initiation, and develop like wt leaves if the timing of their emergence is restored to wt (Conway and Poethig 1997). These observations indicate that during embryogenesis a factor that suppresses vegetative development and/or promotes embryo specic development is expressed in a manner allowing it to affect the prematurely developed leaf primordia. The gene LEAFY COTYLEDON1 (LEC1) is possibly one such factor, as it is expressed in the whole embryo during early embryogenesis. When ectopically expressed during vegetative development LEC1 causes the plant to develop according to embryo specic programs, including the expression of storage proteins and the development of ectopic embryos on vegetative tissue (Lotan et al. 1998). Mutant lec1 cotyledons develop into leaf-like organs, which lack storage proteins and oil bodies present in wt cotyledons, while developing trichomes and vascular characteristics of wt leaves. Mutant lec1 seeds are intolerant to desiccation and, at a low frequency, germinate viviparously (Meinke 1992). However, the embryonic development of lec1 mutants is not fully transformed into the vegetative program, possibly indicating that LEC1 acts partly redundantly with other factors (Lotan et al. 1998).
The development of the cotyledon after the mobilization of storage reserves is similar to that of leaves, although cells of Arabidopsis cotyledons do not divide after the seed is mature (Tsukaya et al. 1994). The mechanisms that regulate the shape of the cotyledon are only partly shared with those of leaves, as an cotyledons develop an abnormally narrow shape (Tsukaya et al. 1994) while the shape of rot3 cotyledons is indistinguishable from that of wt cotyledons (Tsuge et al. 1996).
Sugar sensing
Sugars are important molecules in all organisms, both as carriers of stored chemical energy and as raw materials for the synthesis of other molecules. Sugars are therefore of great importance during all stages of the plant life cycle and it is not surprising that changed sugar concentrations affect the expression of a large number of genes (Koch 1996). Continuous sensing of sugar levels occurs at the level of the individual cells and three major sensing mechanisms/receptors have been suggested to be present in plants, hexokinase (HXK) sensing, hexose transport (HXT) sensing and sucrose (SUT) sensing (Smeekens and Rook 1997).
Hexokinase sensing has been shown to control diverse processes and metabolic pathways in plants such as the sugar induced feedback regulation of photosynthesis and the mobilization of storage reserves from seeds (Smeekens and Rook 1997). The sensor is hexokinase (HXK), the rst enzyme of glycolysis, and phosphorylation of hexoses by HXK induces the enzyme to initiate a signalling cascade (Jang et al.
1997). Two HXK genes have been cloned from Arabidopsis and transgenic plants with enhanced or reduced expression of these genes have conrmed the importance of the proteins in sugar signalling (Jang et al. 1997). HXK can phosphorylate the non-metabolizable sugar analogues 2-deoxy glucose (2dGlc) and mannose, and consequently start signalling in response to the presence of these molecules. HXK cannot phosphorylate two other analogues, 3-O-Methyl Glucose (3-O-mGlc) and 6-deoxy glucose (6dGlc), and these sugars are therefore unable to trigger HXK signalling (Smeekens and Rook 1997). By the use of specic HXK inhibitors Pego et al. (1999) have shown that the signalling properties of HXK, and not the depletion of ATP or phosphates, is responsible for the inhibition of germination in response to these analogues in Arabidopsis.
Sugar analogues that are taken up by the cells but are not phosphorylated by HXK such as 3-O-mGlc and 6dGlc can affect gene expression, but clearly by a second mechanism. Genes encoding extracellular invertase and sucrose synthase are induced by sucrose and glucose in suspension cultures of Chenopodium rubrum cells and this induction can be mimicked by 6dGlc (Roitsch et al. 1995). The sugar and amino acid-induced promoter PAT(33B) can also be induced by 6dGluc and 3-O-mGlc in transgenic Arabidopsis (Martin et al. 1997). The effect of these analogues has been attributed to Hexose transporter (HXT) signalling in plants (Figure 3), in analogy with the function of HXT in yeast (Özcan et al. 1996). HXT proteins in yeast, like HXK in yeast and plants, have dual functions, both transporting hexoses over the membrane and signalling. However, none of the cloned members of the HXT gene family from Arabidopsis has yet been proven to be involved in signalling (reviewed in Buttner and Sauer 2000).
Sucrose
HXT
SUT
HXK
Sucrose Hexose
Hexose
cell membrane
Glycolysis Vacuole
Hexose
Nucleus
Figure 3
Sugar signalling receptors in plants
Schematic drawing of simplified plant cell showing the three suggested sugar receptors (grey) as well as their suggested positions in the cell. The interconversion of sucrose to hexose is indicated.
Sucrose is an important sugar transport and storage molecule in plants and most likely also has a signalling function. Several observations indicate that sucrose can be sensed in plants. The patatin and rolC genes have been shown to be specically induced by sucrose (Jefferson et al. 1990; Wenzler et al. 1989; Yokoyama et al.
1994), and the expression of a bZip transcription factor gene, ATB2, is specically repressed by sucrose (Rook et al. 1998). Since sucrose is readily hydrolysed to glucose and fructose, both inside the cell membrane and in the apoplast, it is difcult to determine if such effects are directly mediated by the sucrose molecule. However, in the above cases, combinations of glucose and fructose are less efcient than sucrose in mediating the response, indicating that sucrose is the specic mediator. A sucrose transporter protein, SUT (Figure 3) is suggested to be involved in signalling but as yet there is no experimental support for this hypothesis. A recently cloned gene from Arabidopsis encodes a protein with sequence similarity to both sucrose transporters and signal transducing cytoplasmic domains of yeast sugar sensors (Barker et al.
2000). This protein might act as a sucrose sensor in plants as it lacks transport activity, like yeast sugar sensors (Barker et al. 2000).
How the signal is transduced from the receptors to the responses in plants is presently unknown. However, in yeast many genes and enzymes are repressed by glucose. This repression has been shown to be controlled by a complex signalling cascade, in which the protein kinase SNF1 plays a central role (Ronne 1995). SNF1- like protein kinases have been isolated from many plant species including rye, barley, tobacco, soybean, potato, Arabidopsis and others (reviewed in Halford and Hardie 1998). Expression of the tobacco SNF1 homologue in yeast causes constitutive expression of a glucose repressible gene (Muranaka et al. 1994) indicating that there is functional homology between the yeast and plant proteins and, possibly, in their signalling cascades. An independent indication of the involvement of SNF1-like proteins in sugar signal transduction is provided by the PRL1 protein. PRL1, when mutated, confers sugar hyper-sensitivity to the plant (Nemeth et al. 1998) possibly by interacting with SNF1-like kinases, as recently shown in vitro (Bhalerao et al.
1999).
Sugar signals are also integrated with signals from other signal transduction pathways in the plants. Mutations identied on the basis of altered sugar responses have also been shown to affect hormone-mediated responses. This is a strong indication of a close connection between the signalling pathways. Crosstalk between sugar and ethylene (Zhou et al. 1998), abscisic acid (Arenas-Huertero et al. 2000;
Huijser et al. 2000; Laby et al. 2000), and cytokinin, ethylene, abscisic acid and auxin (Nemeth et al. 1998) signalling pathways has been demonstrated by this method. It has also been shown that a gene involved in brassinolide biosynthesis is repressed by sugars (Szekeres et al. 1996). Sugar signals form one among many types of input monitored by the highly integrated regulatory network controlling the onset of
owering in Arabidopsis. Sucrose availability to the aerial parts of the plants promotes
owering in the dark (Roldan et al. 1999). In the light, high concentrations of sugars inhibit owering, possibly through repression of the regulatory gene, LEAFY (Otho et al. 1998). Many more examples of crosstalk between sucrose signalling and other signalling pathways are likely to be revealed in the future.
HDZip transcription factors
The pre-history of the HDZip domain
Homeosis is dened as an anomaly where one part of the body develops into the likeness of another part (Bateson 1894). Mutations causing this kind of anomaly have been observed in many organisms including plants and animals and are referred to as homeotic mutations. When the rst genes corresponding to this type of mutants in Drosophila were cloned they were found to share a 180 bp sequence motif that was called the homeobox and the 60 amino acid domain it encoded was named the homeodomain (McGinnis et al. 1984; Scott and Weiner 1984). Homeodomain proteins were rst suggested to act as transcription factors, as the amino acid sequence of the homeodomain has slight similarity to bacterial, viral and yeast transcriptional regulators of the helix-turn-helix type (Laughon and Scott 1984; Shepherd et al.
1984). This was later proven to be correct experimentally, by demonstrations that the factors bind DNA, and affect the transcription levels of nearby genes upon binding, both in vitro (Johnson and Krasnow 1990) and in vivo (Driever and Nüsslein-Volhard 1989). The homeodomain was also found to have a three-dimensional structure similar to that of helix-turn-helix proteins, when the structure of the ANTENNAPEDIA homeodomain was resolved (Qian et al. 1989).
Homeobox genes were also found in other segmented animals and were shown to control processes such as determination of segment identity during embryogenesis.
It was therefore surprising when homeobox genes were found in the non-segmented animal Ceanorhabditis elegans by screening genomic libraries with a degenerate nucleotide pool designed to match all possible codons of the most highly conserved part of previously known homeobox genes (Bürglin et al. 1989). Inspired by Bürglin et al. three different groups independently took the same approach to search for homeobox genes in Arabidopsis (Mattsson et al. 1992; Ruberti et al. 1991; Schena and Davis 1992). All of the groups found homeobox genes, but of a different type to those present in animals. They found genes encoding proteins having two separate domains (the homeodomain and the leucine-zipper domain), known from different types of transcription factors, fused in single proteins. This arrangement was novel and since then has only been found in plant genes, although the complete genomes of other species, including Saccharomyces cerevisiae (Goffeau et al. 1996), Drosophila melanogaster (Adams et al. 2000), Ceanorhabditis elegans (The C.
elegans sequencing consortium 1998), and Homo sapiens have been sequenced (Clinton 2000). Thus, this arrangement, the homeodomain leucine zipper, HDZip, domain is most likely plant specic.
HDZip transcription factors are encoded by a large gene family in Arabidopsis
After isolation of the rst genes encoding Homeodomain leucine zipper (HDZip) proteins many more were isolated from Arabidopsis on the basis of sequence similarity (I; III; V; Carabelli et al. 1993; Lee and Chun 1998; Schena and Davis 1994; Söderman et al. 1994). The HDZip genes of Arabidopsis are grouped into four families HDZip I – IV (Sessa et al. 1994) based on sequence similarity as well as other sequence criteria (summarized in Figure 4). The genes of HDZip I and II appear to share a common origin while members of HDZip III and IV are more distantly related (Chan et al. 1998) and differ in their arrangement in the homeodomain compared to HDZip I and II and to the ANTENNAPEDIA class of animal homeodomains (Figure 4).
The HDZip I and II genes are not clustered on the chromosomes like some animal homeobox genes (Graham et al. 1989). Instead, they are dispersed among all Arabidopsis chromosomes (Figure 5). There is no clear correlation between sequence similarity and chromosomal location, which indicates that most of the gene duplications from which these genes presumably originate are ancient. This is also
HELIX1 HELIX2 HELIX3 LEUCINE ZIPPER
a b c d e f
Class Length of Introns in Extra amino acids mRNA homeodomain* in HDZip domain HDZip I approx. 1500 bases a or none None
HDZip II approx. 1500 bases c and e None
HDZip III approx. 3300 bases d Four (between helix 2 and 3)
and four (between helix 3 and leucine zipper) HDZip IV approx. 3000 bases b and f Seven (between helix 3 and leucine zipper)
* According to the schematic drawing in A
A
B
Figure 4
Summary of differences between the four different classes of HDZip genes in Arabidopsis A, Schematic drawing of the primary sequence of the HDZip domain. The putative helical structures are indicated by boxes. Triangles indicate intron positions.
B, Distinguishing characters of the four different classes of HDZip genes in Arabidopsis.
Adopted from Sessa et al. (1994).
HAT3ATHB12 ATHB20 mi357
mi456
ATHB1
150 125 100 75 50 25
0cM
ATHB13 ATHB23
mi106 mi443
ATHB54
I III
HAT1 ATHB2
ATHB16 ATHB20 mi87
mi123
HAT22
IV
ATHB53ATHB5 ATHB52 HAT2 ATHB33 HAT14 ATHB51 mi121
mi184
V
ATHB21 ATHB6 HAT9
ATHB22 ATHB4 ATHB7 mi421
mi79a
II
ATHB17
Figure 5
Chromosomal locations of HDZip I and II genes
Chromosomal locations of 26 HDZip I and II genes (V; Sessa et al. 1994). The relative positions of ten RFLP markers (mi79a to mi443) are indicated (Liu et al. 1996).
true for HDZip IV genes (Tavares et al. 2000). An ancient origin of the gene family is also supported by the fact that genes encoding HDZip proteins are found in a wide range of land plant species including: tomato (Meissner and Theres 1995), sunower (Chan and Gonzalez 1994), soybean (Moon et al. 1996a), resurrection plant (Frank et al. 1998), poplar (Sterky et al. 1998), carrot (Kawahara et al. 1995; Mattsson 1995), Phalaenopsis (Nadeau et al. 1996), Pimpinella brachycarpa (Moon et al. 1996b), rice (Meijer et al. 1997), maize (Ingram et al. 1999) and ferns (Aso et al. 1999).
HDZip proteins act as dimeric transcription factors
Much of our knowledge of HDZip proteins is deduced from the vast mass of information concerning the biochemical properties of animal homeodomains and leucine-zipper proteins. Homeodomains form a globular structure consisting of three alpha helices. When the factors bind to DNA the third helix is positioned in the major groove. Basic leucine-zipper, bZip, proteins also bind DNA by an alpha-helical structure positioned in the major groove, which is enriched with basic residues. This basic region of bZip proteins extends into a long alpha helix in which every seventh
amino acid is a leucine residue. These residues form a hydrophobic face of the helix that mediates interaction with similar proteins. bZip proteins bind DNA as dimers.
The leucine zipper parts of the protein orientate the basic regions in the major groove of the DNA. HDZip proteins bind DNA as homeodomain proteins and dimerise as bZip proteins in accordance with the fact that they consist of a fusion of these two domains. A simplied model of how HDZip proteins may be oriented when bound to DNA is shown in Figure 6.
Dimers of two identical HDZip proteins have been shown to bind to a pseudo- palindromic DNA sequence, CAATNATTG, consisting of two overlapping identical half-sites (Johannesson et al. 2000; Meijer et al. 2000; Sessa et al. 1993). The homeo- domains of HDZip proteins bind to DNA in a similar fashion to their monomeric counterparts from animals as indicated by mutational analysis of ATHB2. Exchanging residues important for DNA binding of homeodomain proteins in ATHB2 abolishes its DNA binding capacity (Sessa et al. 1997). The bases in the central position of the optimal DNA binding site differ between different HDZip proteins, ATHB2 favours a G/C pair while ATHB1 prefers an A/T pair (Sessa et al. 1993; Sessa et al. 1997). This specicity has been attributed to the amino acids Arg55, Glu46 and Thr56, and unspecied residues outside helix 3 (Sessa et al. 1997). In contrast to ATHB1 and ATHB2, the HDZip I proteins ATHB5, ATHB6 and ATHB16 do not show middle position specicity, and can interact with both sites (Johannesson et al. 2000).
ATHB12 has been demonstrated to interact with a different site (TCAATTAATTGA) composed of the same two half sites but with a different spacing (Chun and Lee 1999). As ATHB5, -6, -12 and –16 are identical to ATHB1 in the positions in helix 3 that are postulated to be important for specicity (positions 46, 55 and 56) (Sessa et al.
1997), the alternative DNA binding preferences of these proteins must be determined by residues somewhere else in the proteins. This hypothesis is also supported by the
nding that some HDZip I and II proteins from rice can interact with both types of site in yeast (Meijer et al. 2000).
The dimerisation specicity of bZip proteins has been extensively studied (Vinson et al. 1993) and the three-dimensional structures of bZip homo- and hetero- dimers bound to DNA (Ellenberger et al. 1992; Glover and Harrison 1995) have been determined. The formation of salt bridges between basic and acidic residues in different proteins is apparently important (Vinson et al. 1993; Glover and Harrison 1995). Assuming that HDZip proteins form dimers like the bZip proteins, HDZip I proteins are very similar in the amino acids facing the putative dimerisation surface, but distinct from HDZip II proteins. In agreement with this notion, all proteins so far tested can form homodimers in solution (Johannesson et al. 2000; Meijer et al.
Figure 6
HDZip domain binding to DNA
Theoretical model of HDZip domain binding to DNA based on the three-dimensional structures of the Drosophila ENGRAILED homeodomain (Kissinger et al. 1990) and the yeast GCN4 leucine zipper motif (Ellenberger et al. 1992). DNA is shown in white and the HDZip domain (secondary structure) in grey. (Johansson, K., Johannesson, H. and Söderman, E. unpublished).
2000; Sessa et al. 1993). Some HDZip proteins from rice and resurrection plants are able to form heterodimers with members of the same class (Frank et al. 1998; Meijer et al. 2000). For example, ATHB5 forms heterodimers with ATHB6 and ATHB16 but not with ATHB1 in vitro (Johannesson et al. 2000). This indicates that there are limitations to the promiscuity of dimerisation among the proteins of the two classes.
Members of the homeodomain and bZip transcription factor families have been shown to either activate or repress (or both) transcription upon DNA binding (Latchman 1998), and this also applies to different HDZip proteins. ATHB1 activates transcription from a promoter with HDZip binding sites, CAAT(A/T)ATTG, (Aoyama et al. 1995) while experiments with ATHB2 indicate that ATHB2 can act as a negative regulator of transcription (Steindler et al. 1999). A fusion construct between the strong
transcriptional activation domain of VP16 from the herpes simplex virus (Dalrymple et al. 1985) and ATHB2 affects transgenic plants in the same way as a reduction in the activity of the wt protein (Steindler et al. 1999). This is a strong indication that ATHB2 acts a transcriptional repressor in wt Arabidopsis. In rice suspension cells the HDZip proteins Oshox1 and –3 (HDZip II) repress transcription, while Oshox4 and –5 (HDZip I) activate it (Meijer et al. 2000). Possibly, HDZip I proteins may generally act as activators and HDZip II proteins as repressors. However, the level of sequence similarity between the different HDZip proteins outside the HDZip domain is low, indicating that different HDZip I and II proteins use different mechanisms to affect transcription.
HDZip proteins are involved in a wide range of processes in plants
The functions of only three HDZip genes; GL2, IFL1 and ANL2, have been identi-
ed by analysis of mutant phenotypes. The glabra2, gl2, mutation causes disturbed trichome and root hair development and the mutant also lacks seed coat mucilage (Di-Cristina et al. 1996; Rerie et al. 1994). The gene INTERFASCICULAR FIBER- LESS1, IFL1, regulates interfascicular ber differentiation (Zhong and Ye 1999) and the gene ANTHOCYANINLESS2, ANL2, affects anthocyanin and root development in Arabidopsis (Kubo et al. 1999). ATHB8 is expressed in procambial cells of the embryo and adult plants and is suggested to be a regulator of vascular development in Arabidopsis (Baima et al. 1995). These four genes belong to HDZip classes III (IFL1 and ATHB8) and IV (GL2 and ANL2).
Altered expression levels of the HDZip II gene ATHB2 (also named HAT4) in Arabidopsis result in phenotypes that suggest it has a role in light signalling (Schena et al. 1993; Steindler et al. 1997). Elevated levels of ATHB2, by means of transgene expression, affect cell expansion in cotyledons and hypocotyls, inhibit secondary growth of the vascular system in roots and inhibit lateral root formation (Steindler et al. 1999). Some of the phenotypic effects can be reversed by application of the plant hormone auxin. Reduced levels of ATHB2 expression by antisense suppression cause reciprocal effects, indicating that the phenotypic effects caused by elevated expression reect the wt function of the gene (Steindler et al. 1999). ATHB2 expression is regulated by far-red-rich light (Carabelli et al. 1993) and the effects caused by altered expression levels of the gene indicate that ATHB2 has a role in shade avoidance responses (Steindler et al. 1999). ATHB4 expression is regulated in the same manner as ATHB2, by far-red-rich light (Carabelli et al. 1993), indicating that ATHB4 has a role similar to that of ATHB2. Two HDZip II genes from resurrection plant, CPHB1
and –2, are regulated by drought and may be involved in gene regulation in response to drought (Frank et al. 1998). Putative homologues of CPHB1 and –2 are present in the Arabidopsis genome and may be involved in similar responses.
Data on the function of HDZip I genes are sparse. Indirect evidence from expression studies indicates that the genes ATHB6, ATHB7 and ATHB12 regulate gene expression in response to both the plant hormone abscisic acid, ABA, and to environmental conditions known to increase endogenous levels of the hormone, such as water deciency (Lee and Chun 1998; Söderman et al. 1999; Söderman et al.
1996). The expression of ATHB7 and ATHB12 is specically and strongly induced by applications of ABA and by treatments known to cause increased endogenous levels of the hormone. The induced expression of ATHB7 is impaired in the ABA- insensitive mutant abi1, suggesting that ATHB7 regulates responses in the ABA signal transduction pathway downstream of the ABI1 gene (Söderman et al. 1996).
ATHB6 is expressed in developing organs and up-regulated by ABA treatments, and is suggested to have a function related to cell division and/or differentiation (Söderman et al. 1999). The expression of the tomato HDZip I gene H52, is up-regulated during pathogen infection. Suppressed expression in transgenic plants results in a miss- regulation of cell death control and regulation of defence genes, which indicates that H52 is involved in limiting the spread of programmed cell death after infection (Mayda et al. 1999). Another tomato HDZip I gene, VaHox1, is expressed in the phloem during secondary growth and may participate in the regulation of processes specic to secondary phases of phloem development (Tornero et al. 1996). The functions of four HDZip I genes, ATHB3, -13, -20 and –23, are discussed in this thesis.
Isolation and characterization of novel HDZip genes
Novel genes distantly related to previously known HDZip I genes
In extensive searches for Arabidopsis HDZip I and II genes in available databases we found, in total, 26 genes. Seven of these were novel and named ATHB21, -22, -40, -51, -52, -53 and 54 (V). We think these 26 genes represent all the HDZip I and II genes in the Arabidopsis genome, as more than 98 % of the estimated Arabidopsis genome has now been sequenced (September 2000). All novel genes encoded a HDZip domain similar to those previously isolated. The deduced sequence of all HDZip domains could be aligned without creating gaps in the alignment, with the exception of ATHB22 (which aligned after creating a gap in the alignment between helixes 1 and 2 in the other sequences, see Figure 7). These amino acids may form a loop in the turn between the two helices, as has been shown for the yeast homeodomain MATα2, which has an insertion of three residues at a similar position (Hall and Johnson 1987). The three-dimensional structure of MATα2 (Wolberger et al. 1991) is very similar to the homeodomains of Drosophila proteins ENGRAILED and ANTENNAPEDIA, indicating that the extra residues in ATHB22 do not affect the structure or DNA-binding properties of the domain.
A homeodomain consensus sequence has been dened on the basis of a compilation of 346 homeodomain sequences from a range of different eukaryotes (Bürglin 1994).
The homeodomains of all HDZip proteins, including the newly identied forms, are highly similar to this homeodomain consensus sequence (Figure 7). The 26 HDZip domains are also very similar at positions 7, 8, 54, and 61. In addition to the conserved homeodomain, all HDZip proteins, including the newly identied HDZip proteins, contain leucine zipper motifs in identical positions, towards the C-terminal from the homeodomain (Figure 7). The sequence similarity between the proteins was signicantly lower in the leucine zipper region compared to the homeodomain.
To investigate if the classication suggested by Sessa et al. (1994) was also applicable to the novel genes, the amino acid sequences of the entire HDZip domain were phylogenetically analysed. The analysis resulted in four equally parsimonious trees with similar branching patterns, one of which is depicted in Figure 8. The four trees all resolve HDZip I and II, indicating that the classication suggested by Sessa et al. (1994) reects the evolutionary history of the gene family. Related genes in the tree have the same or similar intron patterns, further supporting the evolutionary signicance of the depicted tree.
