The Plant Journal (1999) 17(2), 209214

SHORT COMMUNICATION

Diversity and similarity among recognition sequences of Dof transcription factors

Shuichi Yanagisawa1,* and Robert J. Schmidt2 1Department of Life Sciences (Chemistry), Graduate School of Arts and Sciences, The University of Tokyo, Komaba, Meguro, Tokyo 1538902, Japan, and 2Department of Biology and Center for Molecular Genetics, University of California, San Diego, 9500 Gilman Dr, La Jolla, CA 920930116, USA core sequence for precise binding (e.g. Izawa et al., 1993; Schindler et al., 1992; Tabata et al., 1991). It is important to clarify the recognition sequence of each member of a family of transcription factors in order to gain a better appreciation of what each may regulate in vivo. Dof proteins are a family of transcription factors that, so far, are unique to plants. Several Dof proteins have already been identied in both monocots and dicots, but not in yeast and animals (Yanagisawa, 1996). Dof proteins share a unique and highly conserved DNA-binding domain that is composed of 52 amino acid residues. Establishment of the structure of Dof domain still remains. However, it was predicted that a CX2CX21CX2C motif in the Dof domain may form a single zinc nger that is essential for DNA recognition. This was based on experiments showing disruption of DNA binding in mutant Dof proteins bearing substitutions in the cysteine residues, and by metal chelators that can disrupt zinc co-ordination (Chen et al., 1996; de Paolis et al., 1996; Shimofurutani et al., 1998; Yanagisawa, 1995). Four cDNA clones encoding Dof proteins have already been isolated from maize (Vicente-Carbajosa et al., 1997; Yanagisawa, 1995; Yanagisawa and Izui, 1993). Dof1 is a transcriptional activator which shows different activity in greening and etiolated protoplasts, whilst Dof2 appears to be a repressor that can block transactivation of Dof1 (Yanagisawa and Sheen, 1998). It was suggested by in vivo and in vitro binding of Dof1 and enhancement of transcription in vivo, that a target of Dof1 might be the promoter of the gene for C4-type phosphoenolpyruvate carboxylase (C4PEPC) which catalyses a primary carbon xation in maize (Yanagisawa and Sheen, 1998). An endospermspecic Dof protein PBF binds to the prolamin box that is a strong candidate for a cis-element responsive for endosperm-specic gene expression (Vicente-Carbajosa et al., 1997). PBF also interacts with a bZIP protein O2, a genetically dened regulator of zein gene expression in endosperm (Aukerman and Schmidt, 1994). In fact, the binding sites of O2 and PBF are closely located in the promoters of the -zein genes. These data suggest that PBF may be a transcriptional regulator for gene expression in endosperm cells (Vicente-Carbajosa et al., 1997). Indeed, a barely protein, a direct homologue of maize PBF, transactivated an endosperm-specic promoter via interaction with a prolamin box (Mena et al., 1998). The maize Dof3 cDNA is another example of a Dof gene isolated by 209

Summary Dof proteins are a family of transcription factors that share a unique DNA-binding domain. Dof proteins were found recently in association with diverse promoters of plantspecic genes, suggesting various roles of Dof proteins in plants. Through binding site selection experiments using randomly synthesized DNA, the recognition sequences of four maize Dof proteins were systematically analyzed. All selected oligonucleotides contained an AAAG sequence, suggesting that this sequence is the recognition core of Dof proteins. In fact, a single mutation in this sequence abolished binding of all four Dof proteins. Furthermore, the preference of each Dof protein for the sequence anking the core motif was also analyzed using oligonucleotides containing a xed AAAG and random anking sequences. Similar, as well as distinct, anking sequences were observed among the optimal binding sites. Changes in the anking sequences did affect DNA-binding of Dof proteins. Introduction Transcription factors can often be grouped into specic families based on the structural features they have in common. Often, similarity in the DNA-binding domains predicts similar, but non-identical, DNA-binding specicity (e.g. reviews by Gehring et al., 1994; Martin and Paz-Ares, 1997; Menkens et al., 1995; Meshi and Iwabuchi, 1995; Mitchell and Tjian, 1989; Yanagisawa, 1998). For example, all plant basic leucine zipper (bZIP) proteins bind to ACGT core sequence. However, bZIP proteins also require different nucleotide sequences in the anking region of the

Received 4 September 1998; revised 16 November 1998; accepted 19 November 1998. *For correspondence (fax 81 3 5454 6998; e-mail csyanag@komaba.ecc.u-tokyo.ac.jp).

1999 Blackwell Science Ltd

210 Shuichi Yanagisawa et al. homology (Yanagisawa, 1995), although the recognition sequence and function of Dof3 is completely unknown. Dof proteins from dicots have also been reported. An Arabidopsis Dof protein, OBP1, interacted with OBFs (bZIP proteins for stress-response) and enhanced binding of OBFs to DNA (Zhang et al., 1995). The binding site of OBP1 was close to the OBF-binding site in the GST6 gene promoter (Chen et al., 1996). A tobacco protein, BBF, whose cDNA was isolated by in vitro binding to a cis-element for expression of a plant oncogene rolB, also had a Dof domain (de Paolis et al., 1996). A partial cDNA encoding a Dof domain was also isolated from pumpkin (Kisu et al., 1995). The binding of Dof proteins to diverse plant promoters suggests that Dof proteins may be involved in a variety of signal-responsive and/or tissue-specic gene expression in plants. Interestingly, all biological processes that are speculated to be mediated by Dof proteins are plantspecic (e.g. expression of a photosynthetic gene, seed-specic genes and a plant oncogene, and pathogenresponsive gene expression). To better understand the roles of Dof proteins in plants, it is important to begin by assessing the general features of these proteins. We performed systematic analyses of binding-specicity of Dof proteins in vitro as a rst step in delineating target site preferences that could serve as predictors of putative target promoters in vivo. The binding site selection experiment using four Dof proteins of maize revealed that Dof proteins recognize a (T/A)AAAG sequence that is essential for DNA binding. The sequence anking the core motif also inuenced DNA-binding of Dof proteins to a limited extent. Results single mutation in the A at position 2 of the OP1 probe or in the G at position 4 were used. Dof1 bound well to OP1 but did not bind to either Mut-A or Mut-G (Figure 2). Together with our previous observations that Dof1 and Dof2 did not bind to CAAG and TAAG sequences (Yanagisawa and Izui, 1993; unpublished data), these results establish the importance of the AAAG sequence for Dof binding. Similar results were obtained by EMSAs with other Dof proteins (data not shown), suggesting that the AAAG sequence is a common core recognition sequence for Dof proteins.

The preference of each Dof protein in the sequence anking the AAAG core motif
Previously, we have observed that maize Dof1 and Dof2 bound to the AAAAG sequence motif in the C4PEPC gene promoter, but bound weakly to the GAAAG sequence motif and not the CAAAG sequence motif (Yanagisawa and Sheen, 1998). Thus, Dof proteins might display preferences not only for the AAAG core but also for the sequence anking the core motif. In spite of the obvious preference for A or T at position 1 of the Dof proteins (Figure 1), the results of the binding site selection shown in Figure 1 failed to clarify the preferences of Dof proteins for anking sequences due to the following two reasons: (i) Dof proteins frequently selected oligonucleotides containing an AAAG sequence that followed the non-random sequence; and (ii) Dof proteins also frequently selected oligonucleotides with multiple AAAG motifs (Figure 1). In fact, a DNA probe with two tandem repeated AAAG motifs showed about a twofold higher afnity than a probe containing one motif in EMSA (data not shown). Thus, to investigate whether or not Dof proteins show different preferences in the sequence anking the core motif, we carried out a binding site selection with oligonucleotides containing a xed AAAG and short random anking sequences, in which the presence of multiple AAAG was not permitted. After three rounds of selection, selected DNAs were cloned into M13 vector. Mixtures containing single-strand DNAs from 30 independent clones were sequenced. As shown in Figure 3, the preference of each Dof protein for the anking sequence was clearly demonstrated. In the initial selection, Dof1 preferred A or T at position 1, although Dof1 selected T more frequently in this experiment. In EMSA with Dof1 protein, substitution of A for the T at position 1 in the OP1 probe did not signicantly affect the formation of the Dof1-DNA complex (data not shown). In addition, substitution of the A at position 3 in OP1 with a C (which Dof1 occasionally selected in the second selection) increased binding in EMSA, but the effect was less than twofold (data not shown). Thus, the anking sequences can affect the binding but the effects are not as dramatic as observed for substitutions within the core sequence.
Blackwell Science Ltd, The Plant Journal, (1999), 17, 209214

Maize Dof proteins recognize an AAAG sequence motif

To investigate the recognition sequences of Dof proteins systematically, we performed a DNA-binding site selection experiment with four maize recombinant Dof proteins (Dof1, Dof2, Dof3 and PBF). The DNAs to which Dof proteins bound were selected from a library of randomly synthesized DNA. After ve rounds of selection, the sequences of selected DNAs were determined. As shown in Figure 1, all sequences contained at least one AAAG motif, suggesting that this sequence is important for DNAbinding of Dof proteins. To establish the importance of the AAAG motif on DofDNA interaction, EMSA was carried out with a DNA probe (OP1), whose sequence was chosen based on the result of binding site selection. Previously, it has been observed that double mutations at positions 2 and 4 in the AAAG motif abolished binding of several Dof proteins (Vicente-Carbajosa et al., 1997; Yanagisawa and Izui, 1993; Yanagisawa and Sheen, 1998; Zhang et al., 1995). Thus, two other probes (Mut-A and Mut-G) with a

Recognition sequences of Dof proteins 211

Figure 1. Nucleotide sequences of oligonucleotides recovered after ve rounds of selection. The nucleotide sequences were aligned to match the most common sequence, the AAAG motif. In case multiple AAAG motifs were found in an oligonucleotide, all alignments were presented. The sequences of randomly synthesized regions are shown as capital letters and the non-random sequences are indicated in lower case. The frequency of each nucleotide at each position is shown along the bottom of each panel. The last line represents the consensus sequence.

These results support our conclusion that the sequences displayed in Figure 3 reect the optimal binding sequence of each Dof protein. The optimal sequences of Dof1 and Dof2 were barely distinguishable, whereas Dof3 selected more variable anking sequences as optimal. For example, Dof3 preferred C at position 2, whereas others did not. Dof3 did not select G at position 3, but others did. Thus, Dof proteins showed similarity in recognition of a consensus core sequence, but diversity in the preference for the anking sequences. As stated above, nucleotide substitutions in the anking sequence affected the binding afnity only to some degree. Thus, the anking sequence seemed to be exible for DNADof interaction. It is worth noting that the prolamin box (GTGTAAAG) found in the -zein gene promoter was among the optimal sequences selected by PBF, and that the optimal sequences for Dof1 binding were close to the sequences for the Dof1-binding sites identied in the C4PEPC gene promoter (AGAAAAAGTGA, AAAAAAAGTGA, CCAAAAAGGAG and CCAAAAAGAAG).
Blackwell Science Ltd, The Plant Journal, (1999), 17, 209214

Figure 2. EMSA using synthetic DNAs with or without an AAAG motif. Recombinant Dof1 protein (75 ng or 300 ng) was incubated with an OP1 probe or probes with a mutation in the AAAG motif. The Dof1-DNA complexes are indicated.

Discussion We have demonstrated, by binding site selection and EMSAs, that four different maize Dof proteins each recognize the (A/T)AAAG sequence as the core motif. This result is consistent with Dof recognition sequences already identied in putative target promoters. The Arabidopsis

212 Shuichi Yanagisawa et al.

Figure 3. The optimal binding sequences of four maize Dof proteins. The DNA fragments with high afnity to Dof proteins were recovered from a library of oligonucleotides with 4 bp of a random sequence in the 5 anking region and 3 bp of a random sequence in the 3 anking region of a xed AAAG consensus core sequence. The selected DNAs were cloned into the M13 vector and the mixture of ssDNAs from 30 independent clones was sequenced. The sequences of template DNAs are indicated.

OBP1 bound to a CCAAAAGGTGTA sequence in the GST6 gene promoter (Chen et al., 1996). The binding site of tobacco BBF was TAAAGT in the rolB gene promoter (de Paolis et al., 1996). Thus, it may be that all Dof proteins recognize an (A/T)AAAG core sequence. In spite of a common core sequence, the four Dof proteins tested were able to show somewhat different preferences in the anking sequence. However, the effect of anking sequence was limited. In addition, all four Dof proteins were able to recognize some identical sequences (e.g. GTTTAAAGGGG) as optimal. Thus, it is hard to speculate that only the anking sequence determines the specicity of numerous Dof proteins in vivo. It is likely that other factors, such as proteinprotein interactions and post-translational modications, also contribute to specic interactions of Dof proteins with their target sites in vivo. Examples have been documented among animal DNA binding proteins. Phosphorylation of a POU domain transcription factor, Pit-1, modulates its DNA-binding specicity (Kapiloff et al., 1991), and co-operative interaction between two homeodomain proteins selectively increased the afnity for a particular DNA target (Chan et al., 1994). In fact, it has already been reported that some Dof proteins interact with bZIP proteins (Vicente-Carbajosa et al., 1997; Zhang et al., 1995) and HMG1 (Yanagisawa, 1997), and these may modulate proteinDNA interactions. In addition, post-translational mechanisms seemed to regulate the DNA-binding activity of Dof1 differently in greening and etiolated leaves (Yanagisawa and Sheen, 1998). Thus, further analyses on proteinprotein interactions and modications of Dof proteins might provide new insights for specic interaction between Dof proteins and DNA in vivo. The results of Southern blot analyses suggested a minimum of about 1020 Dof-related genes in the maize genome (Yanagisawa and Izui, 1993; unpublished data). Thus, numerous Dof proteins may be present in maize. The delineation of the core consensus recognition sequence of Dof proteins allows us to speculate that some nuclear factors identied previously might be members of the Dof family of proteins. A maize nuclear factor interacting with the P2 promoter region of the maize 19 kDa zein genes was found in endosperm tissues (Maier et al., 1990). Although the binding sequence (AATATCTTTTGCAAATTCGAAATTAATCTT) of this factor was different from the

prolamin box, it does include an AAAAG sequence running 5 to 3 on the opposite strand (underlined). In addition, the binding of this factor was inhibited by a metal chelator, 1, 10-phenanthroline (Maier et al., 1990), a characteristic of Dof proteins. Thus, this factor might be PBF or another Dof protein expressed in maize endosperm cells. Another maize nuclear protein (MNP1) might also be a Dof protein. The consensus sequence (TTAAGAAAAAGGAAACTGGGTTTC) for MNP1, which interacted with the promoter region of the Shrunken gene encoding sucrose synthase at multiple sites (Werr et al., 1988), contains the consensus core binding sequence for Dof proteins established in this study. Analysis of the prolamin box has implicated this sequence motif as a strong candidate for a cis-element regulating endosperm-specic gene expression. However, the prolamin box taken out of the context of its native promoter functioned as a cis-element not only in cultured endosperm cells but also in leaf tissue-derived suspension cells (Ueda et al., 1994). In addition, it has been shown that some sequences in the vicinity of the prolamin box appear to repress the activity of the prolamin box (Ueda et al., 1994). Such examples clearly illustrate the complexity of DNAprotein interactions in vivo. Although our analyses have established a candidate core recognition sequence for Dof proteins, it is apparent that analysis of DofDNA interaction in the context of native promoters in vivo will be required to elucidate the exact means through which precise sequence recognition in vivo is accomplished. Experimental procedures

Construction of expression vectors

A plasmid GST-WT for expression of a fusion protein between glutathione S-transferase (GST) and Dof1 has been described previously (Yanagisawa, 1997). The expression vectors for GSTDof2 and GST-PBF fusion proteins were also described previously (Vicente-Carbajosa et al., 1997; Yanagisawa and Sheen, 1998). An attempt to construct an expression vector for GST-Dof3 fusion protein using a pGEX vector (Pharmacia, Sweden) ended in failure, probably because the lac promoter in pGEX is leaky and expression of Dof3 in E. coli cells may be toxic. Therefore, the expression of Dof3 was carried out under the control of the PL promoter that allows controlled expression of proteins. A DNA fragment of Dof3 cDNA in the gt10 vector was isolated by digestion with EcoRI and NotI, whose sites were located in the linker or downstream Blackwell Science Ltd, The Plant Journal, (1999), 17, 209214

Recognition sequences of Dof proteins 213

of the region encoding a Dof domain, respectively (Yanagisawa, 1995). The fragment was inserted between XbaI and PstI sites in the multiple cloning site of pTrxFus (Invitrogen, USA), an expression vector for thioredoxin fusions, using two oligonucleotides (5GATCCACG-3 and 5-AATTCGTG-3). For this construction, two other synthetic oligonucleotides (5-GGCCGCCCACCATCATCATCATCATTGACTGCA-3, and 5-GTCAATGATGATGATGATGGTGGGC-3) that encoded a 6xHis tag and a stop codon were also used.

DNA-binding site selection experiments

The fusion proteins (approximately 10 g) in bacterial lysates were puried and immobilized on 50 l of the glutathione-sepharose 4B resin (Pharmacia) or TALON metal afnity resin (Clontech, USA) as described previously (Yanagisawa, 1997). Resins were washed with PBS (phosphate-buffered saline) containing 0.1% triton X-100 three times and with PBS twice, and incubated with 40 g of randomly synthesized DNA in 50 l of the binding buffer [100 mM NaCl, 20 mM TrisHCl (pH 8.0)] for 15 min. A library of randomly synthesized DNA was prepared by the annealing of two oligonucleotides (oligo X; 5-GCAAGCTTGTCAGCATGCGGATCN20-AATTCGCAGCTGCAGTCT- AGAC-3 and oligo Z; 5-GTCTAGACTGCAGCTGCGAATT-3) and treatment with Klenow fragment in the presence of [-32P] dCTP (Amersham, UK). After washing the resins with the binding buffer four times, bound DNA was eluted, phenol extracted, ethanol precipitated and amplied by PCR using oligo Y (5-GCAAGCTTGTCAGCATGCGGATC-3) and oligo Z. After ve rounds of selection, bound DNA fragments were cloned into the M13 mp19 vector and sequenced. Selective binding was monitored by radioactivity at the rst round of selection and by agarose gel electrophoresis at subsequent rounds. Similarly, the experiment involving DNA-binding site selection anking the AAAG core was carried out using oligo W (5-GCAAGCTTGTCAGCATGCGGATC-N4-AAAG-N3-AATTCGCAGCTGCAGTCTAGAC-3). After three rounds of selection, selected DNA fragments were cloned between the SphI and PstI sites of the M13 mp19 vector. A mixture of ssDNAs from 30 independent clones was sequenced using a TTH sequencing kit (Toyobo, Japan).

Electrophoretic mobility shift assays (EMSAs)

Preparation of fusion proteins, release of Dof proteins from fusions and EMSAs were carried out as described previously (Yanagisawa, 1997). Synthetic DNA probes had additional GATC sequences attached to each 5 end of the sequences shown in Figure 2. DNA probes were used after end-labeling with the Klenow fragment. DNA-protein complexes were detected by autoradiography, and amounts of complexes were measured by a BAS-1000 Bio-imaging analyzer (Fuji Films, Japan).

Acknowledgements
S.Y. is indebted to Prof. H. Akanuma (University of Tokyo) for his continuous support. This work was supported in part by Grantsin-Aid for Scientic Research from the Ministry of Education, Science and Culture of Japan to S.Y., and a grant from the National Institutes of Health (GM41286) to R.J.S.

References
Aukerman, M.J. and Schmidt, R.J. (1994) Regulation of -zein gene expression during endosperm development. In Results Blackwell Science Ltd, The Plant Journal, (1999), 17, 209214

and Problems in Cell Differentiation Volume 20 (Nover, L., ed.). Heidelberg: Springer-Verlag, pp. 209233. Chan, S.-K., Jaffe, L., Capovilla, M., Botas, J. and Mann, R.S. (1994) The DNA binding specicity of Ultrabithorax is modulated by cooperative interactions with Extradenticle, another homeoprotein. Cell, 78, 603615. Chen, W., Chao, G. and Singh, K.B. (1996) The promoter of an H2O2-inducible, Arabidopsis glutathione S-transferase gene contains closely linked OBF- and OBP1-binding sites. Plant J. 10, 955966. Gehring, W.J., Affolter, M. and Burglin, T. (1994) Homeodomain proteins. Annu. Rev. Biochem. 63, 487526. Izawa, T., Foster, R. and Chua, N.-H. (1993) Plant bZIP protein DNA binding specicity. J. Mol. Biol. 230, 11311144. Kapiloff, M., Farkash, Y., Wegner, M. and Rosenfeld, M.G. (1991) Variable effects of phosphorylation of Pit-1 dictated by the DNA response elements. Science, 253, 786789. Kisu, Y., Esaka, M. and Suzuki, M. (1995) Putative zinc-binding domain of plant transcription factor, AOBP, is related to DNAbinding domains of steroid hormone receptors and GATA-1. Proc. Japan. Acad. 71B, 288292. Maier, U.-G., Grasser, K.D., Haa, M.M. and Feix, G. (1990) Multiple proteins bind to the P2 promoter region of the zein gene pMS1 of maize. Mol. Gen. Genet. 221, 164170. Martin, C. and Paz-Ares, J. (1997) MYB transcription factors in plants. Trends Genet. 13, 6773. Mena, M., Vicente-Carbajosa, J., Schmidt, R.J. and Carbonero, P. (1998) An endosperm-specic DOF protein from barley, highly conserved in wheat, binds to and activates transcription from the prolamin-box of a native -hordein promoter in barley endosperm. Plant J. 16, 5362. Menkens, A.E., Schindler, U. and Cashmore, A.R. (1995) The G-box: a ubiquitous regulatory DNA element in plants bound by the GBF family of bZIP proteins. Trends Biochem. Sci. 13, 506510. Meshi, T. and Iwabuchi, M. (1995) Plant transcription factors. Plant Cell Physiol. 36, 14051420. Mitchell, P.J. and Tjian, R. (1989) Transcriptional regulation in mammalian cells by sequence-specic DNA binding proteins. Science, 245, 371378. de Paolis, A., Sabatini, S., de Pascalis, L., Costantino, P. and Capone, I. (1996) A rolB regulatory factor belongs to a new class of single zinc nger plant proteins. Plant J. 10, 215223. Schindler, U., Beckmann, H. and Cashmore, A.R. (1992) TGA1 and G-box binding factors: two distinct classes of Arabidopsis leucine zipper proteins compete for the G-box-like element TGACGTGG. Plant Cell, 4, 13091319. Shimofurutani, N., Kisu, Y., Suzuki, M. and Esaka, M. (1998) Functional analyses of the Dof domain, a zinc nger DNAbinding domain, in a pumpkin DNA-binding protein AOBP. FEBS Lett. 430, 251256. Tabata, T., Nakayama, T., Mikami, K. and Iwabuchi, M. (1991) HBP1a and HBP-1b: leucine zipper-type transcription factors of wheat. EMBO J. 10, 14591467. Ueda, T., Wang, Z., Pham, N. and Messing, J. (1994) Identication of a transcriptional activator-binding element in the 27kilodalton zein promoter, the 300 element. Mol. Cell. Biol. 14, 43504359. Vicente-Carbajosa, J., Moose, S.P., Parsons, R. and Schmidt, R.J. (1997) A maize zinc-nger protein binds the prolamin box in zein gene promoters and interacts with the basic leucine zipper transcriptional activator Opaque2. Proc. Natl Acad. Sci. USA, 94, 76857690. Werr, W., Springer, B., Schurmann, J. and Bellmann, R. (1988) Multiple interactions between nuclear proteins of Zea mays and

214 Shuichi Yanagisawa et al.

the promoter of the Shrunken gene. Mol. Gen. Genet. 212, 342350. Yanagisawa, S. (1995) A novel DNA binding domain that may form a single zinc nger motif. Nucleic Acid Res. 23, 34033410. Yanagisawa, S. (1996) Dof DNA binding proteins contain a novel zinc nger motif. Trends Plant Sci. 1, 213214. Yanagisawa, S. (1997) Dof DNA-binding domains of plant transcription factors contribute to multiple proteinprotein interactions. Eur. J. Biochem. 250, 403410. Yanagisawa, S. (1988) Transcription factors in plants: physiological functions and regulation of expression. J. Plant Res. 111, 363371. Yanagisawa, S. and Izui, K. (1993) Molecular cloning of two DNA binding proteins of maize that are structurally different but interact with the same sequence motif. J. Biol. Chem. 268, 1602816036. Yanagisawa, S. and Sheen, J. (1998) Involvement of maize Dof zinc nger proteins in tissue-specic and light-regulated gene expression. Plant Cell, 10, 7589. Zhang, B., Chen, W., Foley, R.C., Buttner, M. and Singh, K.B. (1995) Interactions between distinct types of DNA binding proteins enhance binding to ocs element promoter sequences. Plant Cell, 7, 22412252.

Blackwell Science Ltd, The Plant Journal, (1999), 17, 209214