The dwarf phenotype of the Arabidopsis acl5 mutant is suppressed by a mutation in an upstream ORF of a bHLH gene

The polyamines, including putrescine, spermidine and spermine, are ubiquitous components of living organisms and are essential primary metabolites for normal growth and development(Tabor and Tabor, 1984; Heby and Persson, 1990). These low-molecular-weight compounds are positively charged at physiological pH and bind to negatively charged molecules, e.g. nucleic acids, acidic phospholipids and various types of proteins (Cohen,1998). The biosynthesis of putrescine occurs from the processing of ornithine and/or arginine, and is regulated by the enzymes ornithine decarboxylase (ODC) and arginine decarboxylase (ADC), respectively. Subsequent reactions convert putrescine into spermidine and spermine, and these are catalyzed by spermidine and spermine synthases, which add propylamino groups from decarboxylated S-adenosylmethionine, catalyzed by S-adenosylmethionine decarboxylase (SAMDC)(Pegg, 1988).

Studies of higher plants have shown that polyamines play important roles in a wide range of developmental processes, such as embryogenesis, floral development, fruit ripening, senescence, and response to environmental stresses (Evans and Malmberg,1989; Galston and Kaur-Sawhney, 1990; Bouchereau et al., 1999). The acaulis 5 (acl5) mutant of Arabidopsis thaliana is defective in a spermine synthase and exhibits a severe dwarf phenotype, suggesting that spermine is a novel regulator of stem elongation (Hanzawa et al.,2000). Spermine has also been identified as a signal mediator of the defense responses in tobacco(Takahashi et al., 2003; Takahashi et al., 2004). Moreover, transgenic potatoes that express the antisense SAMDC gene under the control of the CaMV 35S promoter with a duplicated enhancer region exhibit stunted growth with highly branched stems and short internodes(Kumar et al., 1996). Overexpression of the oat ADC gene in transgenic tobacco plants results in short internodes, thin stems and leaves, leaf necrosis and short roots (Masgrau et al., 1997). However, the molecular mechanisms by which polyamines control plant growth remain unknown.

The Arabidopsis genome has two genes encoding spermidine synthase, SPDS1 and SPDS2, and two genes encoding spermine synthase, ACL5 and SPMS (Hanzawa et al., 2002; Panicot et al.,2002) (see Fig. S1 in the supplementary material). The spds1 spds2 double loss-of-function mutant shows an embryo lethal phenotype(Imai et al., 2004b). The spms mutant shows no aberrant phenotype, whereas acl5 spmsdouble mutants contain no endogenous spermine but display a dwarf phenotype that is identical to the acl5 single mutant(Imai et al., 2004a). The ACL5 gene is upregulated by auxin, whereas the SPMS gene is responsive to abscisic acid (Hanzawa et al., 2002). Moreover, SPMS interacts with SPDS1 and SPDS2 to form`metabolon' complexes, whereas ACL5 does not interact with either of these proteins (Panicot et al.,2002). Measurement of the polyamine levels in acl5 and spms mutants has also revealed that SPMS is a principal contributor to spermine biosynthesis in vivo(Imai et al., 2004a). These findings thus suggest that ACL5 is specifically required for stem elongation.

To further elucidate the role of ACL5 during stem elongation, we identified extragenic suppressors of the acl5 mutant and designated these as suppressor of acaulis (sac) mutants. We show that the sac51-d mutation disrupts a short upstream open reading frame(uORF) of the SAC51 gene, which encodes a basic helix-loop-helix(bHLH) transcription factor. Our results suggest that ACL5 is involved in the translational control of the SAC51 gene.

The Landsberg erecta (Ler) ecotype of Arabidopsis thaliana was used in all experiments except for those involving axr2-1 (Timpte et al.,1994), dim (Takahashi et al., 1995), spms-1(Imai et al., 2004a) and transgenic lines from the Columbia (Col-0) ecotype. The axr2-1 and gai-1 (Koornneef et al.,1985) mutants were obtained from the Arabidopsis Biological Resource Center. acl5-1 of the Col-0 ecotype was generated by more than seven crosses of the original acl5-1 strain of the Lerecotype (Hanzawa et al., 1997)to wild-type Col-0.

Plants were grown under continuous fluorescent light at 22°C on rock-wool bricks supplemented with vermiculite, or on 0.8% (w/v) agar plates containing MS salts (pH 5.8) and 3% sucrose, after surface sterilization of seeds. Heat-shock treatments of plants were performed as described previously(Matsuhara et al., 2000).

Mutagenesis of the acl5-1 seeds with ethylmethane sulfonate (EMS)was performed as described previously(Takahashi et al., 1992). Briefly, approximately 20,000 acl5-1 seeds were surface-sterilized,hydrated, treated with 0.2% EMS (Sigma, St Louis, MO, USA) for 14 hours and washed extensively. The M1 plants were then divided into 10 pools and self-pollinated. Approximately 2000 M2 seeds from each pool were grown for suppressor screening. Putative suppressors showing recovery from the acl5-1 dwarf phenotype were used for further analysis after being backcrossed three times to acl5-1 of either the Ler or Col-0 ecotypes.

Each of the sac acl5 mutants of the Ler ecotype was crossed to acl5-1 of the Col-0 ecotype. Genomic DNA was extracted from F2 plants showing the acl5 phenotype and analyzed for co-segregation with respect to cleaved amplified polymorphic sequence (CAPS),simple sequence length polymorphism (SSLP) and single nucleotide polymorphism(SNP) markers (Konieczny and Ausubel,1993; Bell and Ecker,1994). These markers were derived from the Arabidopsis Information Resource (TAIR; http://www.arabidopsis.org). The CAPS markers MDC12Dra, MHJ24Dde, MSJ1Eco and MSJ1Hha were developed for mapping the SAC51 locus by using the primer sequences shown in Table 1. DNA sequences were determined from PCR products using a 377 DNA sequencer (Applied Biosystems,Foster City, CA).

sac dim double mutants were identified by initially crossing the sac acl5 mutant with the dim strain. In the F2 generation,plants that are homozygous for the dim allele (dim/dim) and heterozygous for the acl5-1 allele (acl5-1/+) were then selected based on both the dim phenotype and the dCAPS(Neff et al., 1998) of the ACL5 locus. The dCAPS primers used for genotyping the ACL5locus were ACLdF and ACLdR, which produce a single PCR fragment from acl5-1 and two cleaved PCR fragments from ACL5, after XhoI digestion and resolution on an 8% acrylamide gel. All of the selected F2 plants (dim/dim acl5-1/+) were then subjected to growth assays and were further crossed to acl5-1 to determine the SAC genotype. F2 plants that segregated with no progeny of the acl5 phenotype in the F3 generation of this testcross were consequently identified as homozygous sac mutants and their growth data were calculated. The genotypes of the SPMS locus in the sac acl5 spms mutants were confirmed by PCR with SPMS-specific primers (Imai et al., 2004a). For primer sequences, see Table 2.

Total RNA was extracted from 12-day-old seedlings grown on MS agar plates or from flower buds with apical inflorescence meristems of 6-week-old plants. RNA gel blot analyses with EXGT-A1, γ-TIP, ACL, and SPMS, were performed as described previously(Hanzawa et al., 1997; Hanzawa et al., 2002). A gene-specific probe for SAC51 was prepared by PCR from the wild-type genomic DNA using 51F1 and 51R1 primers.

Reverse transcription reactions were done with 1 μg of DNase-treated total RNA and 2.5 pmol oligo(dT) primer in 20-μl reactions using an RNA LA PCR Kit (Takara, Kyoto, Japan), according to the manufacturer's protocol. Quantitative PCR was performed in a DNA Engine Opticon2 System (Bio-Rad,Hercules, CA) using gene-specific primers. Triplicate PCR reactions were averaged. The primers 51F2 and GUSR were used for detecting expression of SAC51-GUS and sac51-d-GUS chimeric transcripts. Relative transcript levels in all samples were normalized using ACTIN8(An et al., 1996). For primer sequences, see Table 2.

For recapitulation of the sac51-d phenotype, a 4.1-kb genomic fragment encompassing from 990 bp upstream of the SAC51 transcription start site to 723 bp downstream of the SAC51 stop codon was amplified from the sac51-d allele with 51F3 and 51R2 primers, digested with BglII, and cloned into the BamHI site of pBI101 (Clontech,Palo Alto, CA), resulting in pSAC51R. For generating the SAC515′-leader deletion construct, the 990-bp SAC51 promoter fragment was amplified with primers 51F4 and 51R3, digested with ClaI and XbaI, and inserted into the ClaI/XbaI-digested pBI101, resulting in pSAC51pro. The SAC51 coding sequence was amplified from genomic DNA using 51F5 and 51R1 primers, subcloned into pGEM-T Easy (Promega, Madison, WI), and further transferred as a SpeI-digested fragment to the XbaI-digested pSAC51pro,resulting in pSAC51Δ5′. For heat shock-inducible SAC51expression, the same SpeI-digested SAC51 fragment was transferred to the heat-shock cassette Ti-vector pTT101(Matsuhara et al., 2000),resulting in pHS-SAC51Δ5′. For heat shock-inducible SPMSexpression, the SPMS cDNA was amplified with SPMSF and SPMSR primers,subcloned into pGEM-T Easy, and further transferred as a SacI fragment to pTT101, resulting in pHS-SPMS. These Ti constructs were used to transform acl5-1 in the Col-0 background. The acl5-1 mutant carrying the HS-ACL5 construct was previously described(Hanzawa et al., 2000).

For GUS expression analysis, the 990-bp SAC51 promoter fragment and the 5′-leader region of either the wild-type SAC51 or sac51-d transcripts were amplified by PCR with 51F3 and 51R4 primers,digested with BglII, and cloned into the BamHI site of pBI101. The construct that contains a point mutation in the SAC51uORF (sac51-C549A-GUS) was generated by a two-step mutagenesis protocol. PCR amplification was first performed using 51F3/mut-C549AR and mut-C549AF/51R4 primer pairs. The PCR products were subjected to a second round of amplification with 51F3 and 51R4, and cloned into pGEM-T Easy. After checking the sequence, the BglII-digested fragment was cloned into pBI101 as described above.

Transformation of Arabidopsis was carried out using the floral dip method (Bechtold and Pelletier,1998) with the Agrobacterium strain C58C1. Transformants were selected in MS agar plates containing 50 μg ml^-1 kanamycin. Independent transgenic lines that segregated 3:1 for the kanamycin-resistance marker in the T2 generation were further selected to isolate progeny that were homozygous for the transgene.

Inflorescence stems of 6-week-old plants were fixed overnight in 50%ethanol, 5% formaldehyde and 5% acetic acid. The samples were then dehydrated through an ethanol series and embedded in Technovit 7100 resin (Heraeus Kulzer, Wehrheim, Germany). Sections (8 μm) were stained with 0.1%Toluidine blue for 15 seconds.

Histochemical and fluorometric GUS assays were performed as described previously (Jefferson et al.,1987). For histochemical analysis, samples were prefixed in 90%acetone at room temperature for 20 minutes. Protein content was determined using the Bradford assay (Bio-Rad) to compare activity to protein units, and each experiment was repeated three times.

Mutations that suppress the dwarf phenotype of the acl5 mutant were identified by screening for tall individuals among M2 plants descended from EMS-mutagenized homozygous acl5-1 seeds. A total of four putative suppressors were isolated from approximately 20,000 M2 plants and designated as suppressor of acaulis (sac) 51-54. After establishing homozygous sac acl5-1plants by self-pollination, we crossed each of them to the wild-type Ler strain and confirmed the segregation of plants showing the acl5 phenotype in each F2 population. Thus, all four sacmutants appeared to represent extragenic suppressors of acl5-1 and not a reversion of the acl5-1 allele. We further backcrossed these sac acl5-1 mutants to the acl5-1 single mutant. F1 plants from the cross between acl5-1 and sac51 acl5-1 were indistinguishable from sac51 acl5-1, whereas those from the cross of acl5-1 to sac52, sac53 or sac54 in the acl5-1 background displayed an intermediate stature when compared with the parental lines (Fig. 1). These results indicate that sac51 is completely dominant, whereas sac52, sac53 and sac54 are semi-dominant. Mapping experiments of each sac locus revealed that SAC51 is located on chromosome V and SAC54 is on chromosome III. SAC52 and SAC53 were mapped to approximately 6.6 cM and 13.3 cM from the SSLP marker nga63 on chromosome I, respectively (see Fig. S1 in the supplementary material). Hereafter, we refer to these dominant or semi-dominant sac alleles as sac51-d to sac54-d.

As shown in Fig. 2A, sac51-d acl5-1 plants are wild type in appearance, whereas sac52-d acl5-1, sac53-d acl5-1 and sac54-d acl5-1 plants have sizes of approximately 68%, 62% and 38% of the wild-type height,respectively. Mature acl5-1 plants do not only display a reduction in their height but also have reduced lengths in their leaf blades, petioles and pedicels, and in their stem diameters(Hanzawa et al., 1997). We found that the restoration of these organ sizes is paralleled by that of the plant height in sac51-d acl5-1 and sac52-d acl5-1 double mutants (Fig. 2A, Table 3). The sac53-dallele is more effective in restoring the stem length (∼61% recovery),than in restoring the stem diameter (∼31% recovery), rendering sac53-d acl5-1 plants more slender than wild-type plants. Interestingly, in sac54-d acl5-1 plants, the pedicel length is fully restored but the stem length is only partially restored (∼39% recovery). Microscopic observations of stem longitudinal sections revealed that the recovery of plant height is attributable to cell lengths in all four classes of sac acl5 mutants (Fig. 2B, Table 3). Our genetic segregation data revealed that all sac single mutants show no obvious phenotype in the presence of the wild-type ACL5 gene (data not shown).

acl5-1 shows an overproliferation of lignified xylem tissues in the stem (Hanzawa et al.,1997). This phenotype is suppressed in sac acl5 mutants,in accordance with the recovery of the plant height(Fig. 2C). The abnormalities in xylem development and lignin accumulation in acl5-1 are reminiscent of those observed in transgenic Arabidopsis plants that constitutively express ATHB8, a gene encoding a class III homeodomain-leucine zipper (HD-Zip III) transcription factor(Baima et al., 2001). The HD-Zip III genes have been implicated in the regulation of vascular differentiation (Baima et al.,1995; Baima et al.,2001; Zhong et al.,1999; Ohashi-Ito and Fukuda,2003), leaf polarity(McConnell et al., 2001) and meristem initiation (Otsuga et al.,2001; Green et al.,2005). We prepared RNA samples from 12-day-old seedlings of wild type, acl5, and four classes of sac acl5 plants, and examined the transcript levels of all five HD-Zip III genes(Prigge et al., 2005) by quantitative real-time RT-PCR. Our results revealed that the expression of each of these genes is upregulated in acl5-1 but is restored to normal levels in sac acl5 mutants, in parallel with the degree of recovery of plant height during the adult stage of growth(Fig. 2D).

A previous study revealed that acl5-1 shows reduced expression levels of EXGT-A1, which encodes the cell wall enzyme endoxyloglucan transferase, and of γ-TIP, which encodes a vacuolar aquaporin,tonoplast intrinsic protein (Hanzawa et al., 1997). To examine the effect of each sac allele on the expression of these genes, northern blot analysis was performed with total RNA isolated from flower buds with apical inflorescence meristems of 6-week-old plants. The levels of both the EXGT-A1 andγ -TIP transcripts were restored in sac acl5 mutants(Fig. 3). We also examined acl5-1 transcript levels in each sac acl5 mutant. In contrast to EXGT-A1 and γ-TIP, the acl5-1transcript level is upregulated in acl5-1, probably as a result of a negative-feedback control of ACL5 expression(Hanzawa et al., 2000). The acl5-1 transcript levels were restored in accordance with the degree of recovery of the plant height in sac acl5 plants(Fig. 3). The recovery in acl5-1 transcript levels was also detected at the seedling stage,prior to the manifestation of morphological phenotypes (data not shown). SPMS transcript levels were not affected in any of the sac acl5 mutants.

To determine whether sac mutations are general suppressors of hormone-related dwarf phenotypes or not, the sac mutants were crossed to an auxin-resistant mutant, axr2, a GA-insensitive mutant, gai, and a BR-requiring mutant, dim. Because axr2-1and gai-1 alleles are dominant, we crossed their respective mutants with sac acl5 and measured the plant height in the F1 generation. sac dim double mutants were identified as homozygotes for both alleles. Our results revealed that none of the sac alleles suppresses the dwarf phenotypes of axr2, gai and dim mutants(Fig. 4A).

acl5-1 spms-1 plants have no detectable levels of endogenous spermine but are morphologically indistinguishable from acl5-1(Imai et al., 2004a). To examine the effect of sac alleles on stem elongation in a background of complete spermine depletion, we generated sac acl5 spms triple mutants. All four of the sac mutants suppressed the acl5phenotype to a similar degree in either the presence or absence of SPMS (Fig. 4B).

We chose the sac51-d allele for further analysis. Fine mapping experiments placed the SAC51 locus within a 60-kb region of the P1 clone MSJ1 (Fig. 5A). All of the genes in this region were amplified by PCR from homozygous sac51-d plants and their sequences were compared with those of the wild-type Ler. We detected a single C- to-T mutation in the At5g64340 gene (Fig. 5B). A genomic fragment encompassing from 990 bp upstream of the transcription start site to 723 bp downstream of the stop codon of At5g64340 was cloned from sac51-d plants and introduced into acl5-1 in the Col-0 background by Agrobacterium-mediated transformation. The resulting transgenic plants displayed the wild-type phenotype in four independent lines (see Fig. S2 in the supplementary material), confirming that At5g64340 is indeed the SAC51gene.

The full-length SAC51 cDNA in the GenBank database is 2,472 bp in length and is separated by three introns. The longest ORF encodes a protein of 348-amino acids with an estimated molecular mass of 37.8 kDa. The SAC51 protein contains a basic helix-loop-helix (bHLH) domain in its C-terminal half(Fig. 5C) and has been designated as AtbHLH142 in the compilation of Toledo-Ortiz et al.(Toledo-Ortiz et al., 2003). SAC51 shows a high sequence similarity over the entire length of the protein(57.3% identity) to AtbHLH143 (At5g09460). The bHLH sequence homologies between SAC51, At5g09460 and three known bHLH proteins in Arabidopsisare shown in Fig. 5D. Most known plant bHLH proteins also exhibit homology outside of the bHLH domain,such as in the MYB-interacting domain of the R proteins(Goff et al., 1992; Abe et al., 1997) or the PAS domain of PIF3 (Kay, 1997). However, such domains are not conserved in SAC51.

The SAC51 cDNA sequence reveals that the SAC51 transcript includes an approximately 870-nucleotide 5′-leader region with five overlapping uORFs (Fig. 5B,C). The sac51-d allele has a C-to-T nucleotide exchange that creates a premature stop codon in the fourth uORF. If the fourth uORF is recognized by a scanning ribosome and translated, it is predicted to encode a 53-amino-acid peptide in wild-type plants but only a 3-amino-acid peptide in sac51-d mutants. The At5g09460 transcript also has a long 5′-leader sequence with five overlapping uORFs. The polypeptide sequence deduced from the fourth uORF of SAC51 shares 69.8% amino acid identity with that of At5g09460.

Northern analysis revealed that SAC51 shows high expression in stems, roots and flowers, but little or no expression in siliques(Fig. 6A). The sac51-dtranscript level was markedly higher in sac51-d acl5-1 and sac51-d seedlings than was the SAC51 transcript level in either wild-type or acl5-1 seedlings(Fig. 6B; data not shown). Other sac mutations did not affect the SAC51 transcript level (Fig. 6B).

The mutation found in the SAC51 uORF suggested that its effects were manifested through the altered translation of the main ORF encoding the bHLH protein. To examine whether deletion of the SAC515′-leader can also serve as a gain-of-function allele and restore the acl5 phenotype, we made the SAC51Δ5′construct, which contains the SAC51 coding sequence fused with its own promoter, but lacks the 5′-leader, and introduced it into acl5-1. We confirmed in all six independent transgenic lines that,although not completely, the construct restored the acl5 dwarf phenotype. The SAC51Δ5′ construct had no effect on the growth of wild-type plants. Moreover, overexpression of SAC51under the control of a heat-shock gene promoter in acl5-1 also partially restored the phenotype in response to heat-shock treatments of the plants (see Fig. S2 in the supplementary material).

We further generated chimeric gene constructs that consist of a 990-bp SAC51 promoter fragment, the entire 5′-leader region of either the wild-type SAC51 or sac51-d transcripts containing five uORFs and three introns, and the GUS reporter gene(Fig. 7A). These two constructs were introduced into wild-type plants and three independent transgenic lines that contained one copy of the transgene were obtained for each construct. In SAC51-GUS plants, GUS activity was detected in young leaves, roots,stems and flowers (Fig. 7B,C),with higher activity in the vascular tissues of both leaves and roots(Fig. 7D,E). In mature embryos,GUS activity was confined to the shoot apex and the root tip(Fig. 7F). In sac51-d-GUS plants, GUS staining was stronger in every tissue examined than in SAC51-GUS plants(Fig. 7B-F). The GUS activity was determined in five seedlings for each individual transformant and was normalized to the GUS transcript level to provide an indication of the translational efficiency of each construct. The sac51-d-GUSconstruct yielded 51.5-fold more GUS activity than the SAC51-GUSwild-type construct. In parallel with sac51-d transcript levels in sac51-d acl5-1 plants (Fig. 6B), however, steady-state levels of the GUS transcript were markedly increased in sac51-d-GUS plants, when compared with those in SAC51-GUS plants. Thus, our results indicate that the sac51-d mutation causes a 2.7-fold increase in GUS translational efficiency, when compared with the wild-type SAC51 construct(Fig. 7G).

To examine whether disrupting the peptide sequence of the SAC51uORF is crucial for suppression of the acl5 phenotype, we generated another GUS construct carrying a C-to-A substitution at the site of the sac51-d mutation (sac51-C549A-GUS). This causes a Gln-to-Lys substitution of the fourth amino acid of the fourth uORF(Fig. 7H). This construct was introduced into wild-type plants and was found to have no obvious effect on transcription and translation of the GUS reporter gene, when compared with the wild-type SAC51 construct(Fig. 7G). The results were reproduced in three independent transformants for each construct.

To address possible regulatory interactions between ACL5 and SAC51, we introduced both SAC51-GUS and sac51-d-GUSconstructs into acl5-1 and sac51-d acl5-1 mutants by crossing experiments, and examined the GUS expression in these mutants. The GUS activity driven by the SAC51-GUS construct in acl5-1 and sac51-d acl5-1 seedlings was about 40% of the levels in the wild-type background, whereas the steady-state levels of the GUS transcript were unaffected in these mutant seedlings. Thus, the GUS translational efficiency in acl5-1 and sac51-d acl5-1was estimated to be 38.8% and 48.2%, respectively, of the levels in the wild-type background (Fig. 7G). The steady-state levels of the GUS transcript from the sac51-d-GUS construct were markedly increased in acl5-1 and sac51-d acl5-1, as in the wild-type background. The GUS activities in these mutants were consequently increased but reached to about 60% of the activities in the wild-type background. These results suggest that ACL5 is required for full activation of SAC51translation.

The fact that ACL5 encodes spermine synthase suggests a role of spermine in the translational activation of SAC51. A previous study reported that the acl5 phenotype is restored by heat-shock treatments of acl5-1 plants carrying the HS-ACL5 construct(Hanzawa et al., 2000). Thus,we finally tested whether or not overexpression of another spermine synthase gene, SPMS, can rescue the acl5 phenotype with the heat shock-inducible HS-SPMS construct. However, the results showed no effect of heat shocks on the growth of acl5-1 plants carrying the HS-SPMS construct, suggesting that ACL5 and SPMStake on different roles in the same cell (see Fig. S2B in the supplementary material).

To elucidate the molecular pathways underlying the stem elongation mediated by the spermine synthase gene ACL5, we have isolated one dominant and three semi-dominant suppressor mutants of the acl5 phenotype,designated sac. These sac mutants are not allelic to each other and suppress the acl5 phenotype to different extents(Fig. 2A, Table 3). The sac51-d acl5-1 mutant is nearly identical in appearance to the wild-type plant,whereas the sac53-d acl5-1 mutant has differential effects in restoring the stem length and the stem diameter. The sac54-d acl5-1plant shows full recovery of the pedicel length but only partial recovery of the stem length. These findings suggest that SAC53 and SAC54act in either an organ- or tissue-specific manner in potential pathways downstream of functional ACL5. Because none of the four sacmutants suppress the dwarf phenotypes of the axr2, gai and dim mutants, we conclude that these sac alleles are not general suppressors of stem growth defects but specifically function in the ACL5-mediated growth regulatory pathways, which can be uncoupled from those of auxin, GA and BR.

Because of the presence of another spermine synthase gene in Arabidopsis, SPMS, all sac acl5 mutants should still be able to produce spermine. This raised the possibility that the spermine produced by SPMS could function in the ACL5-mediated regulatory pathways in sac mutants. However, we found that all four sac mutants can also suppress the acl5 phenotype in an acl5 spms background(Fig. 4B), indicating that the recovery of stem growth does not require spermine in these sacmutants. This result is consistent with the observation that, unlike the HS-ACL5 construct, which can restore the acl5 phenotype in a heat-shock-dependent manner (Hanzawa et al., 2000), the HS-SPMS construct had no effect on the growth of acl5 plants carrying the construct (see Fig. S2B in the supplementary material). Determination of the subcellular localization of ACL5 and SPMS, and/or the identification of an ACL5-interacting protein, will be needed for further understanding of the functional difference between these two spermine synthases.

We observed that the transcript levels of the HD-Zip III class homeobox genes were increased in acl5-1, even at vegetative stages, prior to the visible manifestation of the acl5 phenotype(Fig. 2D). These transcript levels were restored in each sac acl5 mutant, and restoration correlated with the degree of recovery in plant height, suggesting a close association between HD-Zip III gene expression and reduced stem elongation. The Arabidopsis HD-Zip III class genes are mainly expressed in the procambium and developing vascular tissues(Baima et al., 1995; Zhong and Ye, 1999; Ohashi-Ito and Fukuda, 2003; Prigge et al., 2005). Overexpression of a member of this class, ATHB8, promotes xylem formation and results in a strong reduction in plant height(Baima et al., 2001). This may be a consequence of the accelerated differentiation of the primary vasculature, which in turn provokes the anticipated transition to secondary growth. Because acl5-1 stem internodes also have abnormal vascular structures following the overproduction of xylem tissues, the defect in stem elongation in acl5-1 might be attributable to the increased transcript levels of the HD-Zip III class genes, resulting in abnormal vascular differentiation.

A recent study of the thickvein (tkv) mutant, a new allele of ACL5, suggested that a defect in polar auxin transport is responsible for the tkv phenotype that is characterized by the abnormal vasculature in the leaves and stem internodes, and that ACL5/TKV functions in a mechanism that defines the boundaries between veins and the non-vein regions through correct polar auxin transport in provascular cells (Clay and Nelson,2005). ACL5 expression is upregulated by auxin and by the acl5-1 mutation itself (Hanzawa et al., 2000). Moreover, the increased acl5-1 transcript levels are restored in each of the sac mutants, in parallel with the suppression of the dwarf phenotype (Fig. 3). Taken together, these findings suggest that the defect in the polar auxin transport system in the acl5-1 provascular cells causes a local increase in the auxin levels, which in turn results in the overproliferation of provascular cells and in the increased expression of the acl5-1 and HD-Zip III genes. If this is indeed the case, the sac alleles identified would restore the polar auxin transport system in accordance with the recovery of vascular differentiation.

The SAC51 transcript has a long 5′-leader containing five short ORFs upstream of its bHLH protein coding sequence(Fig. 5B). The sac51-ddominant allele was found to have a point mutation in one of these uORFs,which introduces a premature stop codon. Although uORFs occur relatively infrequently in eukaryotic mRNAs, their occurrence is more frequent in growth-related genes such as oncogenes(Kozak, 1987; Kozak, 1991). uORFs are known to impede the translational initiation of their downstream coding ORFs, and this translational repression is released under appropriate conditions(Morris and Geballe, 2000). For example, the yeast GCN4 gene contains four uORFs. These allow the ribosome to reach the start codon of the main ORF only when cells are under amino acid starvation and need to express the GCN4 protein(Hinnebusch, 1990). Such uORF-mediated translational regulation has been investigated in several plant genes including Opaque 2 (Lohmer et al., 1993) and Lc(Wang and Wessler, 1998) in maize, and SAMDC (Hanfrey et al.,2002) and ATB2 (Wiese et al., 2004) in Arabidopsis. Our experiments with transgenic plants carrying GUS reporter fusion constructs revealed that the sac51-d sequence increases both the transcript levels and the efficiency of translation of the GUS reporter gene(Fig. 7G). We also confirmed that the stem growth of acl5-1 was partially restored by transformation with the SAC51 5′-leader deletion construct(SACΔ5′) and the heat shock-inducible SAC51 construct(HS-SAC51Δ5′) (see Fig. S2 in the supplementary material). Thus,we conclude that the sac51-d allele deregulates a posttranscriptional control of the SAC51 gene and that the resulting overproduction of the SAC51 bHLH protein is responsible for the suppression of the acl5phenotype in sac51-d acl5-1. This is consistent with the dominant characteristics of sac51-d. A mutation similar to sac51-dhas been also reported for a uORF of the Arabidopsis ATR1 gene, which encodes a MYB transcription factor (Bender and Fink, 1998). The dominant atr1D allele was identified as an altered tryptophan regulation mutant, with increased expression of a target of ATR1, ASA1, which encodes the anthranilate synthase αsubunit.

Why are the steady-state levels of both sac51-d and sac51-d-GUS transcripts much higher than those of SAC51 and SAC51-GUS transcripts in respective plant lines(Fig. 6A, Fig. 7G)? The most likely possibility is that the sac51-d and sac51-d-GUS transcripts are stabilized by increased translation of their main ORF. Some studies with other plant systems have shown that the increased translation has the potential to protect a transcript from degradation(Sullivan and Green, 1993; Abler and Green, 1996). It is less likely that the sac51-d transcript is stabilized by the alteration its own secondary structure with the C-to-T substitution, because the C-to-A substitution at the site of the sac51-d mutation had no effect on the GUS expression.

Alternatively, the polypeptide encoded by the fourth uORF of SAC51might be involved in destabilizing its own mRNA. The uORF that contains a premature stop codon in sac51-d is the longest one of the five overlapping uORFs of the SAC51 gene and is well conserved among all four members of the bHLH subfamily to which SAC51 belongs(Toledo-Ortiz et al., 2003). The presence of such a conserved uORF suggests that all of these bHLH proteins are under a common posttranscriptional control mechanism. However, we have so far identified no peptide sequences homologous to those of the SAC51uORFs in other plant species in the database. In Neurospora crassa,the arg-2 gene encoding an arginine biosynthetic enzyme contains an evolutionarily conserved uORF. This uORF encodes the 24-amino-acid arginine attenuator peptide (AAP) that confers negative translational regulation in response to increased arginine levels (Luo and Sachs, 1996). At high arginine concentrations, AAP causes ribosomes to stall at the uORF termination codon and create a blockade of additional ribosome scanning. The peptide sequences encoded by uORFs are crucial regulators of some genes, including the cytomegalovirus gpUL4(Alderete et al., 1999) and the mammalian SAMDC (Raney et al.,2000), whereas those of the above mentioned GCN4 gene are relatively unimportant for its function(Hinnebusch, 1996; Hinnebusch, 2000). A more detailed mutagenic analysis of the SAC51 5′-leader sequence will be needed to determine whether the function of the SAC51 uORF is sequence-dependent or not.

One important question is whether ACL5 acts in the control of SAC51 expression. A key finding of our experiments is that the translational efficiency of the SAC51-GUS construct was decreased in acl5-1 and sac51-d acl5-1(Fig. 7G). This implies that the translational efficiency of the SAC51 main ORF depends at least in part on the ACL5 function. As polyamines are known to promote various stages of protein biosynthesis(Tabor and Tabor, 1984), it is possible that spermine is directly involved in the translational control of SAC51. In bacteria, spermidine facilitates the formation of the translation initiation complex (Yoshida et al., 1999). By contrast, in the mammalian SAMDC gene,polyamines may directly participate in an interaction between its uORF-encoded peptide and a constitutive component of the translation machinery, which leads to the inhibition of ribosome activity(Mize and Morris, 2001). We found that the GUS translational efficiency of the sac51-d-GUS gene in the wild-type background was decreased to about 60% in acl5-1 and sac51-d acl5-1 (Fig. 7G), indicating that ACL5 enhances SAC51translation independently of the regulation by its uORFs. Taken together, we propose a model whereby ACL5 acts, either directly or indirectly, as a translational activator of SAC51, and probably of its homologs(Fig. 8). In this model, we also hypothesize that the premature termination of a 53-amino-acid polypeptide encoded by the fourth uORF of SAC51 facilitates the release of the ribosome from the uORF and the reinitiation of translation at the main ORF in sac51-d. The restoration of the acl5-1 transcript level in sac51-d acl5-1 plants, suggests that the negative-feedback control of ACL5 expression does not require spermine itself. This feedback control might involve auxin homeostasis, as discussed above, or downstream components of either SAC51 or its homologs. This model is supported by the histochemical localization of SAC51-GUS expression in transgenic plants (Fig. 7B-F),as SAC51 shows high expression in the shoot apex and vascular tissues of young leaves and roots, a pattern similar to the expression domain of ACL5 (Clay and Nelson,2005).

We could not detect the SAC51 protein in plant extracts by immunoblotting. Attempts to fluorometrically and immunologically detect the SAC51-GFP fusion protein by expressing it under the control of the SAC51 promoter and the sac51-d 5′-leader were also unsuccessful, although this fusion construct restored the acl5 phenotype (data not shown). Because the SAC51 amino terminus contains sequences rich in proline, glutamic acid, serine and threonine (PEST), which are often found in rapidly degraded proteins (Rechsteiner and Rogers,1996), SAC51 might be continuously accumulated and degraded below detectable levels.

In conclusion, our findings shed light on uORF-mediated posttranscriptional control in plant development for the first time. The identification of other SAC genes, and the downstream targets of SAC51, will further elucidate the exact roles of ACL5 and SAC51 during stem internode elongation.

This work was funded by a Grant-in-Aid for Scientific Research (15570027,17370023) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by the Novartis Foundation for the Promotion of Science to T.T. A.I. was supported by a JSPS research fellowship for young scientists.