Mapping sites of gibberellin biosynthesis in the Arabidopsis root tip

Root elongation depends on the action of the gibberellin (GA) growth hormones, which promote cell production in the root meristem and cell expansion in the elongation zone. Sites of GA biosynthesis in the roots of 7-d-old Arabidopsis thaliana seedlings were investigated using tissue-speciﬁc GA inactivation in wild-type (Col-0) or rescue of GA-deﬁcient dwarf mutants. Tissue-speciﬁc GA depletion was achieved by ectopic expression of the GA-inactivating enzyme AtGA2ox2, which is speciﬁc for C 19 -GAs, and AtGA2ox7, which acts on C 20 -GA precursors. In addition, tissue-speciﬁc rescue of ga20ox triple and ga3ox double mutants was shown. Furthermore, GUS reporter lines for major GA20ox, GA3ox and GA2ox genes were used to observe their expression domains in the root. (cid:1) The effects of expressing these constructs on the lengths of the root apical meristem and cortical cells in the elongation zone conﬁrmed that roots are autonomous for GA biosynthesis, which occurs in multiple tissues, with the endodermis a major site of synthesis. (cid:1) The results are consistent with the early stages of GA biosynthesis within the root occurring in the meristematic region and indicate that the penultimate step of GA biosynthesis, GA 20-oxidation, is required in both the meristem and elongation zone.


Introduction
The action of the plant hormone gibberellin (GA) is necessary for normal root growth, although lower GA concentrations are required to achieve maximal rates of root elongation than for the shoot (Tanimoto, 1994(Tanimoto, , 2012. Indeed, supraoptimal GA concentrations can be inhibitory for root growth (Inada & Shimmen, 2000;Coelho et al., 2013). Root elongation proceeds by two mechanisms: cell replication within the root apical meristem (RAM) and cell elongation in the elongation zone (EZ). The RAM is the source of new cells within the root and plays a critical role in defining the lineages of the cell files that result from meristematic divisions. It is located at the distal tip of the root, next to a nondividing group of cells known as the quiescent centre (QC) (Nawy et al., 2005;. As their neighbouring initial cells divide, one of the daughter cells is disconnected from the QC and then differentiates into distinct tissues. The cell files are attached and cannot move relative to one another, consequentially causing a spatial relationship that is indicative of the cell's age, with the youngest cells near the root tip getting progressively older as their distance from the QC increases (Benfey & Scheres, 2000). Tissue-specific suppression of GA signal transduction has demonstrated that this hormone determines both the size of the RAM, that is the number of divisions before the cells exit into the transition zone (Achard et al., 2009;Ubeda-Tomas et al., 2009) and the final cell length achieved within the EZ (Ubeda-Tomas et al., 2008). In both cases, GA acts in the endodermis to allow the coordinated growth of the root cell files (Ubeda-Tomas et al., 2008. Comparison of the growth of cultured roots of wild-type and GA-deficient tomato indicated that roots are autonomous for GA production, which is close to saturating for wild-type root growth (Butcher et al., 1990). It was shown from grafting experiments with Arabidopsis thaliana (Arabidopsis) that the root stocks could restore the growth of GA-deficient scions by supplying the GA precursor GA 12 , while later metabolites are not mobile . GA 12 is biosynthesised from the common diterpene precursor trans-geranylgeranyl diphosphate by the sequential action of the terpene cyclases ent-copalyl diphosphate synthase (CPS) and ent-kaurene synthase and the cytochrome P450 monooxygenases ent-kaurene oxidase and ent-kaurenoic acid oxidase (reviewed by Hedden & Thomas, 2012). GA 12 is converted to the biologically active GA 4 by the 2-oxoglutarate-dependent dioxygenases (2ODDs) GA 20-oxidase (GA20ox) and GA 3-oxidase (GA3ox) (Fig. 1). These last two enzymes are encoded by small gene families, comprising five and four members, respectively, in Arabidopsis (Hedden & Phillips, 2000). Two further clades of 2ODDs deactivate GAs by 2b-hydroxylation, members of one acting primarily on C 19 -GAs, including the biologically active forms, and of the other acting on C 20 -GA precursors, including GA 12 . In Arabidopsis these two GA 2-oxidase (GA2ox) clades contain five and four functional members, respectively (Thomas et al., 1999;Hedden & Phillips, 2000;Schomburg et al., 2003;Lange & Pimenta-Lange, 2020). The 2ODDs are major sites of regulation of GA content, with GA20ox activity limiting the concentration of biologically active GAs in many species, including Arabidopsis (Huang et al., 1998;Coles et al., 1999). Members of the 2ODD families show specificity in terms of regulation and tissue expression, but there is partial redundancy, with three GA20ox enzymes, GA20ox1, -2, -3 (Rieu et al., 2008;Plackett et al., 2012), and the two GA3ox paralogues GA3ox1 and -2 (Mitchum et al., 2006), largely responsible for bioactive GA production required for the proper development of vegetative tissues, including root growth, in Arabidopsis.
Based on transcript abundance, the root tip was indicated to be the main site of GA biosynthesis in the Arabidopsis root, while several GA2ox genes were expressed in the mature region of the root (Dugardeyn et al., 2008). Transcripts for a majority of GAbiosynthetic genes, including CPS, are more abundant in the endodermis than other tissues, while GA20ox1 mRNA accumulates in the cortex. The indication that the endodermis may be a major site of GA biosynthesis is of particular interest given that functional studies revealed that this tissue is also the site of GA action (Ubeda-Tomas et al., 2008). Furthermore, the accumulation of fluorescence in the endodermis when GA-fluorescein probes were applied to Arabidopsis roots suggested the existence of mechanisms to concentrate GAs within this tissue (Shani et al., 2013). Indeed, the transcript for the NPF3 transporter, which was shown to transport the GA 3 fluorescent probe as well as GA 4 and other C 19 -GAs into Xenopus oocytes, accumulates to higher levels in the endodermis than in other tissues (Tal et al., 2015). Rizza et al. (2017) used a fluorescence resonance energy transfer (FRET)-based, nuclear-targeted biosensor (nlsGPS1) to visualise and provide relative quantification of nuclear GA 4 levels within the Arabidopsis root. The output from the sensor indicated a gradient of bioactive GA concentration along the primary root axis increasing with distance from the QC and peaking in the EZ. In contrast with the results reported by Shani et al. (2013), this peak of GA concentration was observed not only in the endodermis, but also in the cortical and epidermal cells.
Here we describe further approaches to define the sites of GA biosynthesis and action within the Arabidopsis root. We use tissue-specific promoters to drive expression of a C 20 -GA2ox, Fig. 1 Biosynthetic pathway from GA 12 to GA 4 and 2b-hydroxy catabolites, indicating the reactions catalysed by AtGA20ox1, AtGA3ox1, AtGA2ox2 and AtGA2ox7.

New Phytologist
AtGA2ox7 and a C 19 -GA2ox, AtGA2ox2 in the Col-0 ecotype, the first expected to act at the site of synthesis, whereas the second could act at the site of synthesis and the site of action. The same promoters are used to express functional AtGA20ox1 in the atga20ox1, atga20ox2, atga20ox3 triple mutant and to express AtGA3ox1 in the atga3ox1, atga3ox2 double mutant, both stacked mutants being extreme dwarfs with short roots (Mitchum et al., 2006;Plackett et al., 2012). The degree of rescue should indicate the location of the reactions lacking in the mutant lines. The extent to which shoot-derived GAs contribute to promotion of root growth is determined by driving expression of the same effector genes from the shoot-specific CHLOROPHYLL A/B-BINDING PROTEIN (CAB) promoter.

Plant material
Arabidopsis thaliana ecotype Col-0 was used as wild-type in all experiments. The ga20ox1 ga20ox2 ga20ox3 triple mutant (Plackett et al., 2012) and the AtGA20ox1::GUS (Hay et al., 2002) and AtGA20ox2::GUS (Frigerio et al., 2006) reporter lines are as previously published. The ga3ox1 ga3ox2 double mutant and AtGA3ox1::GUS and AtGA3ox2::GUS lines are as described in Mitchum et al. (2006) and were provided by Dr Tai-ping Sun, who also provided the ga1-3 (Col-0) mutant line containing the mutation in CPS that was originally found in Ler and was backcrossed to Col-0 six times (Tyler et al., 2004). The reporter lines are translational fusions incorporating promoter and exon and intron sequences. The AtGA2ox6::GUS transcriptional reporter line was provided by Dr Sharyn Perry (Wang et al., 2004).

Growth conditions and replication
Seeds to be grown for seed production, crossing or characterisation were surface sterilised with 10% bleach, rinsed and then imbibed in water or 50 µm GA 4 for 3 d at 4°C in the dark before being sown on Levington's compost. Seeds imbibed in GA 4 were rinsed five times in H 2 O before planting. Plants were grown in a controlled environment in 16-h d (150 µmol m À2 s À1 ) at 24°C : 18°C, day : night temperatures. For measurement of root growth, after imbibition on 50 µm GA 4 and washing, seeds were grown on medium containing 0.5% Gelrite, 1% sucrose, halfstrength Murashige-Skoog medium at pH 5.8 on plates placed at an angle of 10°to the vertical in continuous light (150 µmol m À2 s À1 ). For measurement of root length, two (pseudoreplicate) or three plants of each line were grown on each plate so that the number of plates used was also the biological replication. Hence, there were five or six biological replicates per line for measurement of root length with two or three roots (plants) per biological replicate. For meristem and cell size measurements, a single plant per plate was used and there were five biological replicates. Also for cell size, five (pseudoreplicate) cells were measured for each biological replicate plant. Randomised complete block designs were used with the plates being statistical blocks. For comparison of the effects of each promoter on root length, meristem size and final cortical cell length, the median lines in terms of root length for each promoter were grown together on plates, one line for each promoter per plate with 15 plates as replicates.

Activity of recombinant AtGA2ox2 fusions and AtGA2ox7
To produce plasmid constructs allowing expression of GA2ox2 fused to EYFP, the EYFP coding region was amplified by PCR and cloned into pET-32a (Novagen). Primer sequences are listed in Supporting Information Table S1. The AtGA2ox2 coding region was amplified from a cDNA clone by PCR (Thomas et al., 1999) and cloned into the pET-32a constructs containing EYFP sequences to produce the N-and C-terminal expression constructs N-RB-YFP-AtGA2ox2:pET-32a and C-RB-AtGA2ox2:YFP:pET-32a. AtGA2ox7 was amplified from 35S:: GA2ox7 in pRMA1, provided by Dr R. Amasino (Schomburg et al., 2003), to introduce BamHI and SalI restriction sites and then ligated into pET-32a. The constructs were expressed in E. coli Rosetta 2 (DE3 pLysS) cells (Merck), lysates prepared and GA2ox activity determined by incubating with 14 C-labelled GA substrates (Prof. L. N. Mander, Australian National University, Canberra, ACT, Australia) as previously described (Ward et al., 2010). When necessary, product identity was confirmed by combined gas chromatography-mass spectrometry as described previously (MacMillan et al., 1997), except that samples were run on a MAT95XP mass spectrometer coupled to Trace GC (ThermoElectron) (Rieu et al., 2008) operated in full-scan mode.

Production of transgenic Arabidopsis with tissue-specific expression
The constructs for Arabidopsis transformation were produced using the Gateway cloning system. The YFP:AtGA2ox2 fusion was inserted downstream of the tissue-specific promoters within the Gateway pENTR11 entry plasmid (Marques-Bueno et al., 2016) acquired from the Nottingham stock centre (NASC). AtGA2ox2 was then excised using BamHI and SalI and replaced with AtGA2ox7, AtGA20ox1 and AtGA3ox1. The constructs were transferred to the pGBW7 binary vector following a standard Invitrogen LR reaction. The constructs in pGWB7 were transformed into Agrobacterium tumefaciens strain GV3101, and then into Arabidopsis by the floral dip method (Clough & Bent, 1998). Plants were selected on hygromycin to obtain at least 10 independent T1 lines, segregation analysis was then performed to create at least four stable single insertion homozygous lines for physiological analysis. Stable homozygous, single insertion lines were taken through to the T4 generation before subsequent genotyping and phenotyping.

Plant phenotype analysis
To examine root growth, seedlings were grown on solid MS Gelrite medium for 7 d so that the growth rates of GA-deficient mutants reached maximum (Achard et al., 2009). Seedlings were then mounted on slides in 10 lM propidium iodide solution to Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation New Phytologist (2021) 229: 1521-1534 www.newphytologist.com achieve counterstaining of cell walls around the cortical cells. Confocal microscopy images were acquired using a Leica TCS-SP confocal microscope (Leica, Milton Keynes, UK). EYFP was excited at 488 nm from an argon laser and propidium iodide was visualised at 514 nm. Fluorescence emission for the EYFP was collected between 505 and 530 nm. Fluorescence images were processed using Leica LAS AF LIGHT and Adobe PHOTOSHOP. Subsequently, images were analysed using LAS AF lite, allowing meristem and cell size to be measured (Fig. S1). Measurements of cell length were taken after root hairs became visible. Plants for assessment of shoot phenotype were grown on compost in 5.5-cm pots as described above and photographed 3 and 6 wk after germination.

Histochemical GUS assays and microscopy
To visualise GUS expression, whole plants were placed in a substrate buffer (100 mM sodium phosphate, pH 7, 0.5 mM potassium ferrocyanide, 0.5 mM potassium ferricyanide, and 1 mM 5bromo-4-chloro-3-indolyl glucuronide) and were incubated overnight in the dark at 37°C. Chlorophyll was cleared from stained tissue by incubating in 70% (v/v) ethanol. Whole plants stained with GUS were observed and photographed after decolouration using an Olympus SZ-PT stereomicroscope.
For sectioning, GUS samples were fixed with 2.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.2) for 2.5 h at 4°C. After this, three washes with phosphate buffer were performed. All fixed tissues were dehydrated in a graded series of ethanol, then infiltrated, first with a propylene oxide and then with propylene oxide and Spurr's resin mixture. The samples were then immersed in Spurr's resin overnight at 4°C. Finally, the samples were embedded in Spurr resin. Blocks were sectioned in semithin sections on a Leica EM UC6 ultramicrotome. Semithin sections (1 lm thick) of material prepared for GUS expression were mounted in dibutylphthalate polystyrene xylene and observed with a Leica DMR light microscope. The micrographs were captured using Leica QM500 image analysis software.

Statistical analysis
Analysis of variance was used to test (using the F-test) the overall statistical significance of difference between lines for the root, meristem and cell length data. The analysis took account of the design structure with plates being blocks and sections on plates being plots, with any variation due to pseudoreplicates (roots within plants, or cells within plants) being separated from the biological variation. The standard error of the difference (SED) on the residual degrees of freedom (df) from analysis of variance (ANOVA) was used to calculate the least significant difference (LSD) at the 5% (P = 0.05) and 1% (P = 0.01) levels of significance. The LSDs were used to assign statistical significance to differences between the means of most relevant pairs of lines. Homogeneity of variance over the lines was checked by inspecting plots of residuals and, when required, a natural log transformation was applied with comparisons of means being made on the transformed scale (see figure legends). The GENSTAT (2009, 12 th Edn; VSN International Ltd, Hemel Hempstead, UK) statistical package was used for this analysis.

Expression of GA-metabolism reporter genes in roots
First the expression domains of GA 2ODD genes in the root were investigated using GUS reporters. For AtGA20ox1 and AtGA20ox2, the reporters comprised 1.5 kbp of promoter sequence and the first two exons and introns upstream of GUS (Plackett et al., 2012). GUS staining for the AtGA20ox1 reporter indicated expression in both the meristem and EZ with less expression in the transition zone (Fig. 2a). Within the EZ, expression was confined to the cortical cells (Fig. 2b,c). GUS staining for the AtGA20ox2 reporter was much weaker than for AtGA20ox1, but some staining was detected in the endodermis of the EZ (Fig. 2d). Although we were not able to determine its expression profile, data reported by Dugaredeyn et al. (2008) showed the expression of AtGA20ox3 in roots with a similar distribution to AtGA20ox2. Mitchum et al. (2005) reported the expression from a AtGA3ox1:GUS reporter within the stele throughout the root, except in the tip (meristem and EZ). A root cross-section in the translational fusion reporter line indicates that expression is in the pericycle, particularly at the metaxylem poles, with some staining in the adjacent endodermal cells (Fig. 2e). Mitchum et al. (2005) found a similar expression profile in roots for AtGA3ox2, with additional GUS expression in the EZ, QC and columella, but not in the meristem. In cross-section through the EZ, expression is indicated throughout the stele and also in the endodermis (Fig. 2f).
Expression analysis by real-time quantitative RT-PCR indicates that GA2ox6 is the most highly expressed C 19 -GA2ox gene in roots (Rieu et al., 2008). We examined the expression domain of this GA-inactivation gene within the root using a GA2ox6:: GUS transcriptional reporter line (Wang et al., 2004). Expression was present throughout the differentiated root and absent from the root tip ( Fig. 2g) but, in contrast with GA3ox1, was present in all cell types (Fig. 1h).

Defining promoters for targeted tissue-and stage-specific root expression
To determine the relevance of the observed expression domains in the root, the sites of GA production necessary for promotion of root growth were defined by tissue-specific GA inactivation or mutant rescue. The promoters selected to drive tissue-specific expression are listed in Table 1. A 2-kb promoter sequence from the GLABRA2 (GL2) gene was used to confine expression to epidermal cells of the root. Expression occurs in cells outside a periclinal cortical cell wall and extends from the meristem to the root hair zone, with most intense expression in the differentiation zone (Masucci et al., 1996). The GL2 expression domain is defined by a 500-bp fragment located between positions À840 and À1340 in the 5 0 region of the gene (Hung et al., 1998).

New Phytologist
Co2 expression within the root is confined to the meristematic region of the cortex, its root expression domain conferred by a 500-bp promoter sequence immediately upstream of the 5 0 end of the open reading frame (Heidstra et al., 2004). Publicly available data https://www.arabidopsis.org/servlets/TairObject?type= locus&name=At1g62500 indicate the expression of Co2 also in numerous aerial organs, including juvenile rosette leaves and inflorescence, with peaks of expression within the carpel and ovary. Co2 is highly expressed at the shoot apex during the vegetative, inflorescence and transition stages, more specifically within the rib meristem and peripheral zone of the shoot apical meristem (SAM). The Co2 promoter was first used by Mustroph et al. (2009) as part of a cell sorting experiment investigating plants' response to hypoxia.
The CoR promoter was shown by fluorescence-activated cell sorting of root cells to be expressed specifically within the cortex of the EZ . Outside of the root it is expressed in germinating seeds and seedlings, leaves, reproductive organs and in the rib meristem and peripheral region of the SAM. A 500-bp sequence upstream of the start codon was used as this has been shown by green fluorescent protein (GFP)-targeted expression to confirm its expression pattern in the root (Vasesva et al., 2018).
SCARECROW (SCR) encodes a GRAS transcription factor that is expressed in cortex/endodermal initials of the developing embryo and then solely in the endodermis (Di Laurenzio et al., 1996). A 2-kb 5 0 upstream sequence of the SCR gene was used to drive expression, which as well as in the root, occurs in vegetative and reproductive aerial organs in the cell layer surrounding the vascular tissues .
We used a 2-kb promoter sequence upstream of the start codon of SHORTROOT (SHR), which is expressed in the stele . Its product moves from the stele into the adjacent endodermis to regulate SCR expression (Nakajima et al., 2001). It is expressed also in aerial tissues, being strongly expressed in the peripheral zone of the SAM (Yadav et al., 2009) and in the leaf vasculature (Dhondt et al., 2010).
In order to assess the contribution of shoot-derived GAs to root growth, a 1892-bp 5 0 upstream sequence of the CAB gene was used to drive expression in green tissues in the absence of significant root expression. The CAB gene responds to many of the developmental signals required for chloroplast development and green tissue specification (Puente et al., 1996). Experiments using CAB1::GUS fusions have shown that in wild-type Arabidopsis CAB is first expressed in the cotyledons in response to light, and after further growth the expression is confined to the photosynthetic tissue (Yadav et al., 2002).

Tissue-specific GA inactivation
To inactivate C 19 -GAs and C 20 -GAs by 2b-hydroxylation, Arabidopsis Col-0 was transformed with AtGA2ox2 and AtGA2ox7 constructs, respectively (see Fig. 1). Their cDNAs were  transformed as 5 0 translational fusions with YFP in constructs containing the promoters listed in the previous section (Table S2). AtGA2ox2 has been previously shown by in vitro enzyme assays with bacterial expressed protein to be specific for C 19 -GAs (Thomas et al., 1999). To confirm that YFP fusions retained enzyme activity, the activity of N-and C-fusions produced in E. coli was compared with that of AtGA2ox2 against [17-14 C]GA 9 and [17-14 C]GA 4 as substrates. HPLC analysis of the products demonstrated that both YFP fusion proteins were active with both substrates (Fig. S2). Therefore, it was decided to use N-terminal YFP fusions for each of the effector proteins. The assumption that the fusion proteins retained enzymatic function is supported by their expression from the CAB promoter, which produced the expected shoot phenotype (to be described later). The specificity of AtGA2ox7 was confirmed by incubating recombinant enzyme with 14 C-labelled substrates: the C 20 -GAs, GA 12 and GA 24 and the C 19 -GAs, GA 9 and GA 4 (Fig. S3). While both C 20 -GAs were 2b-hydroxylated by the enzyme, we could detect no activity against GA 9 and GA 4. GA 12 was metabolised by AtGA2ox7 to GA 110 (2b-hydroxyGA 12 ), and GA 24 was converted to a product consistent with the previously undescribed 2b-hydroxyGA 24 . Three to six homozygous T3 lines with single inserts were selected for each transformed construct and the presence of the transgene was confirmed by PCR. The expression domains within the root were confirmed by imaging YFP fluorescence using confocal microscopy (shown for representative lines for each promoter in Fig. S4). Root lengths, as well as meristem and final cortical cell lengths (see Fig. S1) were determined in seedlings 7 d after germination and compared with those for Col-0 and the highly GA-deficient ga1-3 mutant. The data for three representative YFP-AtGA2ox2-expressing lines for each construct are shown in Fig. S5, while the median lines in terms of root length for each AtGA2ox2-containing construct are compared in Fig. 3. With the exception of CAB, each promoter driving YFP-AtGA2ox2 resulted in a reduction in root length, with SCR having the most severe effect and SHR producing a relatively small reduction. SCR, Co2 and GL2, but not the other promoters, reduced meristem size, while SCR and particularly CoR reduced final cell length. The equivalent data for the promoter:YFP-AtGA2ox7 lines are presented in Fig. S6 and Fig. 4. The effect of these constructs on root growth was less severe than for the AtGA2ox2 constructs, but some reduction in root length was obtained with SCR, CoR, Co2 and GL2. Significant (P < 0.01, LSD) reductions in meristem size compared with Col-0 were found with the SCR, Co2 and GL2 promoters, while SCR, CoR and GL2 also produced shorter cortical cells.
The effect of the ectopic gene expression on growth of the shoot was assessed at 3 and 6 wk after sowing on compost. Results for the AtGA2ox2 effector are shown in Fig. S7. Most severe dwarfism was obtained with the CAB promoter, while some lines transformed with CoR:YFP-GA2ox2 also exhibited substantial dwarfism. Mild dwarfism was noted for plants harbouring genes driven by the other promoters, except for SHR:YFP-GA2ox2 plants, which were identical to Col-0. In contrast with AtGA2ox2, expression of AtGA2ox7 produced substantial shoot dwarfism regardless of the promoter (Fig. S8). In this case, CAB and Co2 were the most effective promoters, with SCR and GL2 intermediate, and CoR and SHR producing milder dwarfism.
Rescue of the ga20ox1, ga20ox2, ga20ox3 triple mutant Loss of AtGA20ox1, AtGA20ox2 and AtGA20ox3, which are three of the five GA20ox genes in Arabidopsis, produces a severely dwarfed, sterile plant with reduced root length and seeds which do not germinate without application of GA (Plackett et al., 2012). The triple mutant was transformed with genes in which YFP-AtGA20ox1 was downstream of the same tissue-specific Seedlings were grown on vertical plates and lengths calculated using IMAGEJ software. † Indicates that the transgenic lines are statistically significantly different from ga1-3 and * indicates that they are statistically significantly different from Col-0. 12 plants for each line were measured for root length with two plants (roots) per biological replicate (n = 6), five plants for each line were measured for meristem and cell size (n = 5), with five cells per biological replicate being measured. *, P < 0.05; **, P < 0.01; †, P < 0.05; † †, P < 0.01 according to least significant difference (LSD) values (see Supporting Information Notes S1 for means, LSDs and df). Error bars indicate AE SE.
New Phytologist (2021) 229: 1521-1534 Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation www.newphytologist.com Research New Phytologist promoters described above. The root expression domains were confirmed in representative plants for each construct by confocal microscopy (Fig. S9). Total root lengths, meristem and final cortical cell lengths were determined 7 d after germination and compared with Col-0 and the triple ga20ox mutant. The data for three lines are presented in Fig. S10, and for the median lines for each construct, Col-0 and triple ga20ox mutant in Fig. 5. Complete rescue of root growth, including meristem size and cortical cell length, was obtained with the SCR promoter, while Co2, GL2 and CAB provided partial rescue of root length. In the case of Co2 and GL2, both meristem and cell lengths were promoted, while the CAB promoter increased only cell length. No rescue was obtained with the SHR or CoR promoter. The failure of CoR: YFP-AtGA20ox1 to rescue cell length, whereas there was slight rescue by the Co2 promoter, suggested that cell elongation required GA 20-oxidation in the meristem. Indeed, full rescue of cell elongation and root growth was obtained by combining CoR: YFP-AtGA20ox1 with Co2:YFP-AtGA20ox1 in the ga20ox triple mutant (Fig. 6).
The effect of the transgenes on growth of the shoots is illustrated in Fig. S11, in which the lines are compared with Col-0 and GA-deficient triple ga20ox and ga1-3 mutants. Growth of plants expressing CAB:YFP-GA20ox1 was slightly greater than for the wild-type, and the SCR and Co2 promoters gave almost full mutant rescue. Expression from the GL2 promoter also gave substantial rescue, but the CoR and SHR promoters driving YFP-GA20ox1 produced no growth stimulation in the triple ga20ox mutant.

Rescue of the ga3ox1 ga3ox2 double mutant
The same tissue-specific promoters were used to drive expression of YFP:AtGA3ox1 in the ga3ox1 ga3ox2 double mutant; the spatial distribution of YFP in roots of the transgenic lines was determined by confocal microscopy (Fig. S12). Apart from in their expected root expression domains, the GL2 and CAB promoters also produced fluorescence in the columella. Total root lengths, meristem and final cortical cell lengths for three lines are shown in Fig. S13, and the median plants for each construct are compared with Col-0 and the double ga3ox mutant in Fig. 7. All promoters driving YFP:AtGA3ox1 rescued root length to some extent, with roots in the SHR, SCR, CoR and GL2 promoter lines not being significantly (P > 0.05, LSD) shorter than the Col-0 wild-type. In the Co2 and CAB lines there was partial rescue of root length. Meristem length in the SHR, SCR and Co2 lines was not significantly different (P > 0.05, LSD) from that in Col-0, while there was partial rescue of meristem size in the CoR and GL2 lines, but none in the CAB lines. There was substantial rescue of final cortical cell length with all promoters, with only the Co2 line producing cells significantly (P < 0.05, LSD) shorter than those in Col-0.
Similar to what was observed for root growth, the constructs fully restored growth of the shoots, regardless of which promoter was used (Fig. S14). Rizza et al. (2017), using a nuclear-targeted biosensor nlsGPS1, reported a gradient of bioactive GA from the root tip to the EZ that was exaggerated when GA 4 was applied to the root. This indicates the existence of a transport system that concentrates bioactive GAs in this zone. The results of Shani et al. (2013) from application of fluorescein-conjugated GA 4 and GA 3 probes to the Arabidopsis root support this conclusion, although, while Shani et al. (2013) found accumulation of the probe within the endodermis, the results of Rizza et al. (2017) suggest wider distribution of GA within the EZ. However, these findings are not in agreement with the model proposed by Band et al. (2012), which predicts that GA concentration is diluted as cells expand, as a potential mechanism to explain growth cessation. Furthermore, Seedlings were grown on vertical plates and lengths calculated using IMAGEJ software. † Indicates that the transgenic lines are statistically significantly different from ga1-3 and * indicates that they are statistically significantly different from Col-0. Twelve plants for each line were measured for root length with two plants (roots) per biological replicate (n = 6), five plants for each line were measured for meristem and cell size (n = 5), with five cells per biological replicate being measured. *, P < 0.05; **, P < 0.01; †, P < 0.05; † †, P < 0.01 according to least significant difference (LSD) values (see Supporting Information Notes S1 for means, LSDs and df). Error bars indicate AE SE. it has remained unclear as to whether bioactive GAs are synthesised in situ or are translocated to the sites of action from other tissues as final products or as precursors. The existence of a concentrating mechanism within the root, as suggested by Rizza et al. (2017) and Shani et al (2013), is compatible with movement to the EZ from another site. In the present study to investigate the sites of GA biosynthesis in the root, the promoters chosen to provide cellspecific GA inactivation or mutant rescue were based on their established expression domains within the root, which were confirmed by expressing YFP fusions. This multipronged approach should enhance the robustness of the results, which could be influenced by differences in the expression levels of the ectopic genes compared with those of the native GA-biosynthesis genes.

Discussion
The selected promoters are also active in aerial tissues and several of the constructs modified shoot as well as root growth. To determine the extent to which root growth is influenced by the shoot, the same effector open reading frames were expressed from the shoot-specific CAB promoter. While expression of the GA inactivation AtGA2ox2 and AtGA2ox7 cDNAs from CAB resulted in severe dwarfing of the shoot, these constructs had no effect on root growth, confirming that roots can be autonomous for GA production. It has been reported that application of the GA-biosynthesis inhibitor uniconazole to shoots of Arabidopsis seedlings stimulates root elongation and upregulates the feedback-regulated GA20ox and GA3ox genes in roots (Bidadi et al., 2009). Bidadi et al. (2009) argued that, under standard conditions, root growth is promoted partially by GA or a precursor transported from the shoot, but that, in the absence of this source of promotion, the plant can enhance in situ GA production to (more than) compensate. Grafting experiments have shown that Final cortical cell length. Seedlings were grown on vertical plates and lengths calculated using IMAGEJ software. † Indicates that the transgenic lines are significantly different from and atga20ox1,2,3 and * indicates that they are significantly different from Col-0. 12 plants for each line were measured for root length with two plants (roots) per biological replicate (n = 6), five plants for each line were measured for meristem and cell size (n = 5), with five cells per biological replicate being measured. *, P < 0.05; **, P < 0.01; †, P < 0.05; † †, P < 0.01 according to least significant difference (LSD) values (see Supporting Information Notes S1 for means, LSDs and df). Error bars indicate AE SE.

Research
New Phytologist the shoot can provide the GA precursor GA 12 to rescue root growth in GA mutants with lesions early in the GA-biosynthetic pathway, while later intermediates and the active hormone are apparently not transported . Transformation of the ga20ox triple mutant with CAB::YFP:AtGA20ox1 and of the ga3ox double mutant with CAB::YFP:AtGA3ox1 gave Comparison of the lengths of the primary root, proximal meristem and final cortical cells in the cross and parent lines compared with those in Col-0 and the triple atga20ox1,2,3 mutant. Seedlings were grown on vertical plates and lengths calculated using IMAGEJ. † Indicates that the transgenic lines are significantly different from ga1-3 and * indicates that they are significantly different from Col-0. 12 plants for each line were measured for root length with two plants (roots) per biological replicate (n = 6), five plants for each line were measured for meristem and cell size (n = 5), with five cells per biological replicate being measured. *, P < 0.05; **, P < 0.01; † †, P < 0.01 according to least significant difference (LSD) values (see Supporting Information Notes S1 for means, LSDs and df). Error bars indicate AE SE.
Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation New Phytologist (2021) 229: 1521-1534 www.newphytologist.com partial recovery of root length and cell length, but not meristem size, indicating that shoot-derived C 19 -GAs can make a small contribution to root growth, even if GA 12 is the main form transported from the shoots. Gibberellin signalling is required for normal cell production in the meristem (Achard et al., 2009;Ubeda-Tomas et al., 2009) and for cell elongation (Ubeda-Tomas et al., 2008), with the endodermis identified as the major site of GA action. Our results based on rescue of the triple ga20ox mutant indicate that GA 20oxidation in the endodermis (SCR promoter), cortex (Co2) or epidermis (GL2) is sufficient to stimulate root growth, with the endodermal activity having the largest influence (Fig. 5). On the basis of meta-analysis of gene expression data, Dugardeyn et al. (2008) concluded that many of the genes that influence GA production in the Arabidopsis root are most highly expressed in the endodermis, but also have expression in the cortex. Expression of AtGA2ox2 and AtGA2ox7 from the Co2 promoter, which is active in cortical cells of the meristem, reduced meristem size, but not cell length, while expression of these effectors from CoR, which promotes expression in the cortex of the EZ, reduced cell length,

Research
New Phytologist but not meristem size. This is consistent with GA being required in both the meristem and EZ for normal root growth. It also suggests that the cortex is an important site of GA biosynthesis. However, while AtGA20ox1 expression from Co2 gave partial rescue of both meristem and cell size in the triple ga20ox mutant, expression of AtGA20ox1 from CoR gave no rescue, even of cortical cell length. Based on GUS staining in lines expressing the AtGA20ox1 reporter, the endogenous gene is expressed specifically in cortical cells of the EZ as well as in the meristem (Fig. 2a-c). The unexpected lack of rescue by CoR::YFP: AtGA20ox1 may indicate insufficient availability of its C 20 -GA substrates in the EZ, which may be provided by the meristem. This was tested by crossing CoR::YFP:AtGA20ox1 with Co2:: YFP:AtGA20ox1 to restore GA20ox activity in the cortex of the meristem and EZ. Full rescue of the root phenotype in the cross confirmed that GA biosynthesis, and potentially activity within the meristem is necessary for cell elongation. If the meristem provides C 20 -GA substrate for GA 20-oxidation in the EZ cortex, this would not be GA 12 , which would be depleted by GA20ox activity, when in fact Co2::YFP:AtGA20ox1 restored some cell elongation, although this could also be explained by provision of active GA to the EZ. Full rescue of cell elongation was obtained with CoR::YFP:AtGA3ox1 in the ga3ox1/ga3ox2 background, indicating the presence of its substrate GA 9 in the cortex and that its product GA 4 is able to move to the site of action in the endodermis. It is also possible that GA action in the cortex enables some cell expansion. This rescue of cell elongation by CoR::YFP: AtGA3ox1 indicates that GA biosynthesis and/or action in the meristem is not required for cell elongation, although CoR::YFP: AtGA3ox1 also provided partial rescue of meristem size suggesting movement of GA 4 from the EZ to the meristem. It is necessary to consider the upregulation of GA20ox genes in the GAdeficient mutants due to feedback regulation (Phillips et al., 1995) that could increase GA 9 production and availability in cells where it is normally limiting.
On the basis of reporter gene expression, Mitchum et al. (2006) showed that AtGA3ox1 is expressed in the mature, nonelongating region of the root and our results with the same reporter lines indicate that this expression is in the pericycle cells of the stele. Using an equivalent reporter gene, Bidadi et al. (2009) obtained GUS staining in the EZ and root tip after shootapplied uniconazole. This indicates that the gene is expressed in these regions, albeit normally at very low levels. AtGA3ox2 is also expressed in the mature root, but in addition is expressed in the EZ, QC and columella, but apparently not in the meristem (Mitchum et al., 2006). Taken with the nlsGPS1 reporter data, this could indicate that small quantities of GA are activated around the QC providing a low concentration of GA required for cell division. Then, as the cells reach the transition zone, the higher expression of GA3ox1 increases the amount of bioactive GA to promote cell elongation. We show that within the EZ, AtGA3ox2 is expressed in the stele and endodermis, with some expression in cortical cells. This expression domain is consistent with the complete rescue of root growth in ga3ox1 ga3ox2 by expression of AtGA3ox1, which encodes a functionally identical enzyme to AtGA3ox2, from the SHR (stele), SCR (endodermis) and CoR (cortex) promoters. Expression in the epidermis from the GL2 promoter also gave strong rescue of root growth, although there is no indication that GA3ox genes are normally expressed in this cell file. This rescue demonstrates that these tissues have access to the GA3ox substrate GA 9 and that the product GA 4 is accessible to the site of action in the endodermis. The absence of AtGA3ox1 causes a small reduction in root growth, whereas loss of AtGA3ox2 has no effect (Mitchum et al., 2006). This would suggest that GA 4 produced from GA 9 in the stele makes a contribution to promotion of root growth. While there is evidence for the long-distance movement of GA 12 , but not of GA 9 , from the shoot to roots , there may be some limited transport of GA 9 from the shoot or it may be produced in the phloem. However, SHR::YFP:AtGA20ox1 failed completely to rescue root growth in the triple ga20ox mutant, indicating that the SHR expression domain does not overlap with those of AtGA20ox1, AtGA20ox2 or AtGA20ox3. Full rescue of the growth of aerial organs in the ga3ox1 ga3ox2 mutant was obtained with all promoters driving AtGA3ox1 expression, while no rescue of the triple ga20ox mutant occurred when AtGA20ox1 was expressed from SHR or CoR. The other promoters gave different degrees of triple ga20ox rescue, with CAB and SCR proving the most effective. It would appear that the cells expressing SHR and CoR have limited access to GA 12 or other C 20 -GA intermediates, whereas GA 9 is freely available in all the expression domains tested. It is also possible that GA 9 is not released from these cells or, less likely, that there is cell-specific instability of GA20ox1. Expression of the C 20 -GA-inactivation cDNA AtGA2ox7 was generally more effective than that of the C 19 -GA2ox, AtGA2ox2 at reducing shoot growth with almost all promoters, except for CAB, which gave strong dwarfism with both effector genes. This may reflect the strength and broad expression domain of the CAB promoter in shoots. For AtGA2ox7 expression, the SHR and CoR promoters were the least effective, consistent with the GA20ox mutant rescue and reinforcing the conclusion that the stele and cortex of the EZ have least access to C 20 -GAs. By contrast, with the exception of CAB, the CoR promoter driving YFP:AtGA2ox2 was the most dwarfing, with SCR having some effect. If there is high mobility between tissues of GA 9 and GA 4 , as suggested by the ga3ox mutant rescue, then, providing their formation occurs in more than one tissue type, inactivation restricted to specific tissues will be ineffective. The dwarfing obtained with CoR and SCR promoters could reflect the fact that they are active at, or close to, the site of GA action.
In conclusion, the results, which are summarised in Fig. 8, support previous indications that roots and shoots are autonomous for GA production and that GA is required for cell division in the root meristem and elongation in the EZ. The cortex and endodermis are important sites of GA biosynthesis, but GA biosynthesis (GA20ox activity) and/or cell elongation in the cortex of the EZ are dependent on GA production and/or action in the meristem. The meristem may also supply GA and precursors to the EZ. While transport of C 20 -GAs or earlier precursors of GA biosynthesis between root tissues is restricted, the rescue of root growth by AtGA3ox1 in the ga3ox double mutant indicates efficient movement of C 19 -GAs between tissues. Partial restoration of cell elongation, but not cell division, by shoot expression of GA20ox1 or GA3ox1 in the respective mutants indicate some movement of C 19 -GAs from shoot to root, despite GA 12 being the major transport form .

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.      Comparison of the primary root parameters for three lines expressing YFP-AtGA2ox7 from tissue-specific promoters in Col-0 with those of Col-0 and ga1-3 in 7-d-old seedlings.

Fig. S7
The effect on shoot growth in Col-0 of expressing AtGA2ox2 from tissue-specific promoters. Four to six independent transformants are shown for each construct with Col-0 and ga1-3.

Fig. S8
The effect on shoot growth in Col-0 of expressing AtGA2ox7 from tissue-specific promoters. Four to six indepen-dent transformants are shown for each construct with Col-0 and ga1-3. Fig. S10 Comparison of the primary root parameters for three lines expressing YFP-AtGA20ox1 from tissue-specific promoters in the ga20ox1 ga20ox2 ga20ox3 triple mutant with those of Col-0 and the triple ga20ox mutant in 7-d-old seedlings.

Fig. S11
The degree of shoot growth rescue of the ga20ox1 ga20ox2 ga20ox3 triple mutant from expressing AtGA20ox1 from tissue-specific promoters.   Notes S1 Means, least significant differences (LSDs) and degrees of freedom (df) for the data presented in Figs 3-7.
Table S1 Primer sequences for amplification of cDNAs and promoters.