Seeds of A. thaliana ecotype Col-0 as well as the T-DNA insertion mutants lcd1 (N584529) and ppdk (N616157) were obtained from the Nottingham Arabidopsis Stock Centre (NASC). In addition, the lines cue1-6 (Col-0 background), cue1-1 and cue1-3 were used (Streatfield et al., 1999). The line pOCA108 (Bensheim background) served as a control plant for cue1-1 and cue1-3 as described in Li et al. (1995). Mutant and wild-type plants overexpressing the PPT from Brassica oleracea (var. cauli) or the PPDK from Flaveria trinervia were described in detail by Voll et al. (2003). Plants were either germinated and grown on soil or on sterile half-strength ½Murashige-Skoog (½MS) agar (horizontally and vertically) in the absence of sucrose for 3–5 weeks. In both cases light/dark cycles of 16/8 h, a day/night temperature of 22/18°C, and a photon flux density (PFD) of 150 μmol·m⁻²·s⁻¹ on leaf level were applied in a temperature controlled growth chamber equipped with a mixture of fluorescence tubes (Osram L58W/11-860, L58W/30, L58W/76 and L58W/77). Plant material for RNA extraction was harvested after 5 h in the light. Phytohormones were sterile-filtrated before mixing with the ½MS agar during the cooling process and applied at final concentrations of 1, 10 or 50 μM.

In order to test the role of long distance transport for the development of the cue1 phenotype, reciprocal grafts were generated between wild-type and cue1 mutant lines as described by Christmann et al. (2007).

Arabidopsis thaliana was transformed with recombinant Agrobacterium tumefaciens by the floral dip method (Bent, 2006). Plants were grown on soil in a temperature controlled greenhouse. In order to increase the number of flowers, the first inflorescences were removed to induce growth of numerous secondary inflorescences. Recombinant A. tumefaciens was grown in YEB medium in the presence of the respective antibiotic until an OD₆₀₀ of 0.8–1.0 was approached, concentrated by centrifugation at 2500× g for 15 min, and the sediment re-suspended in infiltration medium supplemented with Silwet L-77 (Lehle Seeds, Round Rock, USA) at a ratio of 1:2000. Before transformation already developed siliques were removed.

It was intended to transiently inhibit PPT1 expression with the aid of an inducible promoter system. The alcR-alcA promoter system (Syngenta; Caddick et al., 1998) is based on the dissociation of the alc repressor (transcription factor) (alcR) from the alcA promoter in the presence of ethanol or acetaldehyde. For transient expression, the PPT1RNAi constructs is controlled by the alcA promoter, whereas the alcR transcription factor is driven by the CaMV 35S promoter. For cloning pGreenII containing BASTA® resistance was used.

Stable transformed plants were grown for 2–3 weeks on ½MS agar and were then transferred to fresh ½MS agar supplemented with 0.015% (v/v) EtOH. The appearance of the reticulated leaf phenotype was correlated with the abundance of PPT1 transcripts. For withdrawal of EtOH, the transformants were again transferred to fresh ½MS agar plates.

The analysis of free amino acid contents in A. thaliana leaves and roots was essentially carried out as outlined in Prabhakar et al. (2010) by reversed phase HPLC using a Nucleodur 100-5 C18 ec column (Macherey and Nagel). The samples were precolumn derivatised with o-phthalaldehyde (Lindroth and Mopper, 1979).

Rosette leaves and roots of WT and cue1 plants were harvested at stage 1.03 (12 DAS) from 3 biological replicates. The data on cytokinin and auxin metabolites represent the mean ± SE of six or two measurements for leaves and roots, respectively. Extraction and measurements of auxin and cytokinin levels were performed as described before (Novák et al., 2008, 2012).

For all histochemical analyses, the Eclipse E800 microscope (Nikon) equipped with differential interference contrast and fluorescence optics was used. Images were captured using a 1-CCD color video camera (KY-F1030; JVC) operated by the DISKUS software package (Technisches Buero Hilgers). For the microscopic analyses of the numbers of sub-epidermal palisade parenchyma cells, leaves were incubated in 0.1% Triton X-100 and centrifuged as described by Horiguchi et al. (2006). Mesophyll cells in an area of 0.05 mm² were counted with the aid of Image J (http://rsb.info.nih.gov/ij/).

RNA was extracted according to (Logemann et al., 1987). After treatment with DNA-free DNase (Ambion), oligo(dt)-primed cDNA of total RNA was synthesized using the Bioscript reverse transcriptase (Bioline). PCR using Taq polymerase (Promega) according to Mullis et al. (1986) at between 25 and 30 cycles. For the analysis of PPT1 transcript abundance the following primer combination was used; PPT1for, AGTGCTGAGGAAGGTGATAAC and PPT1rev, AGCCATCAGGGCGAGAGACA. As a loading control Actin2 was used, with Actin2for, TGTACGCCAGTGGTCCTACAACC and Actin2rev, GAAGCAAGAATGGAACCACCG. Amplified DNA was separated on 0.8–2.0% agarose gels in the presence of Tris-acetate buffer and 5 μg·ml⁻¹ ethidium bromide.

The data are expressed as mean values ± standard error of the mean (SE) of the indicated number of independent measurements. Significant differences between two data sets were analyzed using the Welch-test, which allows for unequal relative errors between two groups of measurements assuming that a Gauss distribution is applicable (Welch, 1947). Significant differences between more than two data sets were analyzed using single site ANOVA combined with the post-hoc Tukey-Kramer test, which allows the comparison of unequal sample sizes and identifies those pairs of values, which are significantly different from each other (Ludbrook, 1998). For data plotting and fitting, SigmaPlot10.0 for Windows (SPSS Inc.) was used.

The deficiency in PPT1 results in a complex phenotype of the cue1 mutant alleles, which can be divided into four separate sub-phenotypes. These are (i) growth retardation of the shoot (Figures 1D,F) compared to the control or wild-type plants (pOCA108, Figure 1A; Col-0, Figure 1B), (ii) a reticulate leaf structure based on smaller mesophyll cells and chloroplasts therein. (iii) More pronounced serrations at the margins of the leaves lead to an altered overall leaf shape (Figure 1D; see Byrne, 2012). Finally (iv) the roots are stunted, most prominently in the cue1-1 allele (Figure 1G; Bensheim [pOCA] background) compared to cue1-6 (Figures 1H,I; Col-0 background). In contrast, the lcd1-1 mutant shares only the reticulate leaf phenotype with cue1 (Figure 1C). Size of the rosettes, shape of the leaves and root length (Figure 1J) were very similar to the wild type. The cue1-1 mutant was crossed with lcd1-1 resulting in the cue1-1/lcd1-1 double mutant (Figure 1E) that lacked any additional phenotype compared to the cue1-1 single mutant (Figure 1D).

In previous reports it was hypothesized that the leaf phenotype of cue1 results from a deficiency in metabolic signals, which might be produced in the roots, where only PPT1, but not PPT2, is highly expressed, and transferred to the shoots by long distance transport via the phloem or xylem (Knappe et al., 2003b; Voll et al., 2003). We addressed this hypothesis by reversed grafting experiments using young scions or root stocks of cue1-1 and pOCA, the control plant of cue1-1 (Figure 2). About 30% of the grafted plants survived and were further analyzed. To our surprise, cue1-1 scions grafted on pOCA root stocks (cue1-1|pOCA plants), still showed smaller, reticulated leaves (Figures 2B,F) and were hence comparable to the controls (Figures 2A,F,H; i.e., cue1-1|cue1-1 plants). Likewise, pOCA roots developed normally despite of the fact that cue1 scions were grafted on the pOCA root stocks (Figures 2B,F). Moreover, root length and branching was comparable to the control, i.e., pOCA|pOCA plants (Figure 2C). Furthermore, if pOCA scions were grafted on cue1 root stocks, both shoots and roots retained their phenotypes, i.e., the shoots still looked wild-type like, whereas the cue1 roots remained stunted in growth (Figures 2D,H). These experiments demonstrate that long distance communication between roots and shoots (and vice versa) with regard to the individual phenotypes does not occur. Hence, it is likely that both phenotypes are separate and independent phenomena, based on the deficiency in PPT1 in the individual organs.

Previously it has been shown that cue1 plants grown on a medium supplemented with a cocktail of AAA loose both the growth and reticulate leaf phenotypes. This has been considered as a proof for the essential role of PPT1 in PEP supply for the shikimate pathway in plastids. In order to examine their potential in rescuing the cue1 phenotype, it was intended to apply a variety of compounds to the cue1 mutant in a transient system. First it was tested whether mature mutant plants could be transiently rescued by AAA. For this purpose two-week-old plants, with a fully developed leaf phenotype were transferred to ½MS agar supplemented with a cocktail of all three AAA (Figure 3). Indeed, the reticulate leaf phenotype disappeared within seven days of feeding. However, growth of rosette leaves could not be restored within this short time period. Moreover, after back-transfer of the rescued plants to ½MS agar devoid of AAA, the reticulate leaf phenotype re-appeared within three to five days, most prominently in newly developed leaves (not shown).

In a second approach, PPT1 expression was transiently repressed using a PPT1RNAi construct under the control of the EtOH-inducible alcR/alcA promoter (Figure 4B). As a positive control, Col-0 wild-type plants expressing the PPT1RNAi construct driven by the constitutive CaMV 35S promoter (Figure 4A) showed all aspects of the cue1 phenotype (Figures 4D,E) compared to the wild type (Figure 4C). In the absence of the inductor EtOH, Col-0 plants carrying the inducible PPT1RNAi construct were indistinguishable from the wild type (not shown). An RT-PCR time series following the induction of the alcR/alcA- PPT1RNAi revealed that PPT1 expression declined severely 8 h after application of with 0.015% EtOH (Figure 4F). The minimum expression level was attained after approximately 16 h. Following the withdrawal of EtOH, the PPT1 transcript abundance recovered slowly and reached a substantial level after approximately five days. The altered PPT1 transcript abundance was accompanied by phenotypic changes of the rosette leaves. The reticulate leaf phenotype appeared in the alcR/alcA-PPT1RNAi plants within seven days after induction with EtOH (Figure 4H), and disappeared four days after withdrawal of EtOH (Figure 4J) compared to the respective control plants (Figures 4G,I).

Both experiments, the transient application of AAA and the RNAi-dependent suppression or induction of the PPT1 gene, clearly demonstrate that the cue1 reticulate leaf phenotype is reversible.

In previous reports it has been speculated that the cue1 phenotype might be caused by a depletion of neolignans such as DCA (or its glucoside DCG) deriving from the phenylpropanoid metabolism (Streatfield et al., 1999; Voll et al., 2003). Both the aglycon and glucoside are commercially not available. For our experiments we used DCA and DCG preparations, which were originally synthesized by David Lynn (Orr and Lynn, 1992). Aliquots of the same preparations were also used in the tobacco system by Tamagnone et al. (1998a,b). The cue1 plants grown on ½MS agar plates were transferred to the same medium supplemented with either 20 μM DCG or DCA, but there was no rescue detectable within one week. However, once the roots had been removed and the cue1 shoots were placed with their cut surfaces directly on the agar, the reticulate leaf phenotype disappeared within seven to nine days after feeding with DCG (Figure 5A), but not with DCA (not shown). The phenotype re-appeared once new roots had been formed at the cut surface of the shoot. Thus, the newly formed roots blocked the direct uptake of DCG into the plant. After removal of these new roots and back-transfer of the shoots to DCG containing medium, the leaf phenotype disappeared again within about one week (not shown).

This result was quite promising, as it showed that a downstream product of the phenylpropanoid pathway was capable of rescuing the cue1 phenotype. Feeding of other lignans like lariciresinol und pinoresinol (the aglycons) at a concentration of 20 μM had no rescuing effect on the cue1 leaf phenotype in the presence or absence of the roots (not shown). In addition, the direct precursor of lignin and lignan biosynthesis, coniferyl alcohol also failed to rescue the reticulate leaf phenotype of cue1.

In view of the reported cytokinin-like effects of DCG in the tobacco callus culture system (Binns et al., 1987; Lynn et al., 1987; Teutonico et al., 1991), we applied trans-zeatin (tZ) to intact cue1 plants and to cue1 deprived of the roots. This highly active cytokinin was capable of rescuing the cue1 phenotype within five to seven days after transfer of the plant to a tZ containing medium. Figure 5B shows the effect of tZ feeding at a concentration of 1 μM on the reticulated leaf phenotype of cue1-1. The phenotype almost completely disappeared after seven days of feeding. However, a prolonged feeding with tZ (e.g., for 11 days) led to a general deterioration of the phenotype and bleaching of rosette leaves indicating that cytokinins can have toxic effects when applied at elevated concentrations for a prolonged time period (Hung et al., 2004; Rosar et al., 2012).

Other phytohormones such as the naturally occurring auxin indole acetic acid (IAA), the synthetic auxin naphtyl acetic acid (NAA), abscisic acid (ABA), gibberellic acid (GA3), the ethylene precursor 1-aminocyclopropane-1-carboxylic acid (ACC), salicylic acid (SA), or methyl jasmonate (MeJA) were ineffective in rescuing the leaf phenotypes of cue1 or lcd1 (Supplemental Figure 1).

The A. thaliana lcd1 mutant shares the reticulate leaf phenotype with cue1, which is also based on a smaller mesophyll cell size. Although the lcd1 mutant is not allelic to cue1 (González-Bayón et al., 2006) and the function of the LCD1 protein is still unknown, we tested whether either a cocktail of AAA or tZ were also capable of rescuing lcd1. In contrast to previous analyses, which were based on phenotypic comparisons at a macroscopic scale (Figures 6A–C), we investigated the effect of AAA and tZ on the level of mesophyll cell density. In A. thaliana leaves the mesophyll cell density was highly variable depending on the age or the developmental stage of the leaf (Figure 6D), a fact that had been considered in this study when comparing treated and untreated wild-type or mutant plants. Compared to the wild type the density of the mesophyll cells of cue1-6 slowly increased within seven days upon feeding of AAA or tZ and was almost doubled seven days after onset of feeding (Figure 6). The increase of the mesophyll cell density of cue1 was related to a macroscopic rescue of the reticulate leaf phenotype (i.e., the leaves appeared uniformly green [Figures 6A–C]). Moreover, the rescue of the mesophyll cell number in cue1 occurred in a concentration-dependent manner (Supplemental Figures 2, 3). Hence, both substance classes, AAA and cytokinins, were capable of rescuing the underlying cause for the reticulate leaf phenotype in cue1, but not in lcd1. In the presence of tZ, newly emerging leaves of both cue1 and lcd1-1 appeared to be greener compared to the control condition (Figures 6A,C), which masked the lack of effect on the cell density in lcd1-1 (Figure 6H and Supplemental Figures 3K,L).

Having established that both AAA and tZ could rescue the diminished cell density in cue1, but not in lcd1, we addressed the question as to whether a similar rescue could be brought about for the stunted root phenotype of cue1. If, for instance, AAA availability were limiting for growth of cue1 roots, feeding with a cocktail of AAA should also relief growth retardation of this organ. Strikingly, root growth was further inhibited by AAA application not only in cue1 but also in wild-type and lcd1-1 plants (Figure 7A). This inhibitory effect was even more pronounced when Phe, Tyr, and Trp were applied individually. The biosynthesis of amino acids, in particular of AAA, is highly regulated and target of feedback-inhibition by end-products (Tzin and Galili, 2010). Feeding of Trp resulted in a general arrest of root growth in all plant lines (Figures 7B–G). These data show that the application of AAA at millimolar concentrations specifically rescued the leaf, but not the root phenotype of cue1. It has been shown earlier that high concentrations of AAA can have an inhibitory effect on plant growth (e.g., Voll et al., 2004). However, even 10-fold lower concentrations of AAA (as well as tZ) lacked any promoting effect of the root length of cue1 (not shown). Hence, a deficiency in AAA as a cause for the stunted root phenotype of cue1 appears to be unlikely.

Moreover, the effect of individual applications of Phe, Tyr, and Trp on the shoot phenotypes of cue1 and lcd1 was also analyzed. In contrast to the AAA cocktail, individual feeding of Phe, Tyr, or Trp (2 mM each) failed to rescue the reticulate leaf phenotype of cue1 (Supplemental Figure 4).

It has been shown that active cytokinins, such as benzyladenine (BA), inhibit root-, but promote shoot growth (Cary et al., 1995; Werner et al., 2001, 2003; Howell et al., 2003). Short-term feeding of 1 μM tZ led to a substantial inhibition of root growth in all plant lines tested (i.e., pOCA, Col-0, cue1-1, cue1-3, cue1-6 and lcd1-1, Figure 8). For cue1-1 the growth inhibition of the roots was even stronger than in the corresponding control plant pOCA. As for the AAA treatment, feeding with tZ had opposite effects on the shoot and root phenotype of cue1. In Figure 8B, the effect of tZ on root growth is also shown for the weak cue1 allele cue1-3.

As shown in Supplemental Figure 1, phytohormones such as NAA, ABA, GA₃, ACC (as ethylene precursor), SA or MeJA neither rescued the reticulate leaf phenotype of cue1 nor of lcd1. Here we have studied the effect of phytohormone application on root growth of various cue1 alleles and lcd1-1 compared to the respective control plants. In Supplemental Figure 5, the root phenotypes are shown, whereas Figure 9 shows the evaluation of the experiment. The application of NAA at a concentration of 1 μM resulted in an inhibition of root growth in wild-type plants. In lcd1-1 inhibition of root growth was less pronounced and it was absent in the cue1-1 and cue1-6 allele, suggesting that the responsiveness to auxins is diminished in cue1 roots. The effects of ABA, ACC, GA₃, and MeJA appeared to be more heterogeneous and responded to the different A. thaliana ecotypes rather than to the individual mutations (Figure 9). For instance, ABA promoted root growth in Col-0, but inhibited root growth in pOCA (i.e., Bensheim). Likewise GA₃ inhibited root growth in cue1-6, but had no effect on cue1-1. In the presence of SA, root growth of both cue1 alleles was severely inhibited.

The above data show that a rescue of the stunted root growth could neither be accomplished by the application of AAA nor by phytohormones.

It has already been shown that constitutive overexpression of a heterologous PPT (from cauliflower buds; BoPPT) or of a PPDK from Flaveria trinervia (FtPPDK) rescued the reticulate leaf phenotype of cue1 (Voll et al., 2003). In A. thaliana, PPDK is a single copy gene showing highest expression in pollen and moderate expressed in other cells and tissues. In leaves PPDK is induced during senescence (eFP browser, Winter et al., 2002) and by cytokinin treatment (Brenner et al., 2005). In order to investigate the impact of endogenous PPDK on the phenotype of cue1, the ppdk-1 knockout mutant was isolated, established as homozygous line and crossed to cue1-1. The ppdk-1 single mutant was phenotypically indistinguishable from the wild type (Figure 10A) and the leaves of the double mutant exhibited the cue1-1 phenotype (Figures 10C,D). However, root growth of the cue1-1/ppdk-1 double mutant exhibited a slight recovery (Figures 10B,G). Feeding of tZ to the cue1 single mutant and the cue1-1/ppdk-1 double mutant rescued in both cases the reticulate leaf phenotype (Figure 10E), indicating that the tZ effect is independent from PPDK induction by cytokinins. Moreover, the additional inhibition of root growth in the presence of tZ was identical in cue1-1 and the cue1-1/ppdk-1 double mutant (Figure 10F).

Furthermore, root growth was analyzed in the already existing BoPPT and FtPPDK overexpressing lines (Voll et al., 2003) in the wild-type and cue1 backgrounds (Figure 10G). The constitutive overexpression of BoPPT inhibited root growth in the Col-0 background by 40%, whereas, in the cue1-6 background, stunted root growth was relieved from about 60% in cue1-6 to 20% in cue1-6:BoPPT compared to the wild type. In contrast, overexpression of FtPPDK in the Col-0 background lacked any significant effect on root growth neither in Col-0 nor in cue1-6. The roots of the cue1-6:FtPPDK line remained stunted, with a length of about 40% of the wild-type. These data show that, apart from the complementation with BoPPT only the absence of PPDK in the cue1-1/ppdk-1 double mutant had a recovering effect on root growth of cue1.

Feeding of tZ rescued the reticulate leaf phenotype of cue1, but inhibited root growth in wild-type, cue1, and lcd1 plants. Hence, it was investigated whether endogenous levels of cytokinins are altered in both roots and leaves of cue1 compared to the wild type. The contents of numerous cytokinin and auxin metabolites have been analyzed in roots and leaves (Table 1). As most striking result, there was a significant more than two-fold increase of the active cytokinin tZ in roots of cue1 compared to the wild-type control. The endogenous levels of almost all other isoprenoid cytokinin metabolites were also elevated. This most likely relates to increased cytokinin biosynthesis as the levels of key cytokinin nucleotides (c/tZRMP, iPRMP, DHZRMP) are several times higher in cue1 plants. Dihydrozeatin (DHZ) and aromatic cytokinins were present at detectable levels.

Despite the fact that neither tZ nor cis-zeatin (cZ) contents showed any significant differences in leaves, there were interesting changes in most less active and inactive cytokinin metabolites, such as ribosides und glucosides (Table 1). Of the 20 glycosides, 13 were significantly increased in leaves of cue1 compared to the control. The aromatic cytokinin metabolites (BAP, oT, mT) were usually present at detectable levels and did not show any clear trends (Table 1). Strikingly, although the total contents of cytokinin metabolites were almost doubled in leaves and roots of cue1 compared to the wild type, the amount of active cytokinins was not dramatically changed.

For auxins, there were no pronounced differences in the contents of indole-3-acetic acid (IAA), its glutamate conjugate and the IAA precursor indole-3-acetamide (IAM) (Table 1).

As the amount of tZ was almost doubled in roots of cue1 compared to the wild type, it is conceivable that growth retardation of the roots is a consequence of tZ accumulation. It has been shown that the ectopic overexpression of a cytokinin oxidase (CKX; CKO) is capable of increasing root growth in transgenic A. thaliana (Werner et al., 2003). Two transgenic A. thaliana lines overexpressing CKX have been crossed with cue1 and established as stable lines (Figures 11C,D). Roots of the overexpressing lines CKX1 and CKX3 (Figure 11A) were longer and more branched than roots from the wild-type or from cue1-6, whereas the shoots were retarded in growth (Figure 11B). The rosettes of CKX3/cue1-1 (Figure 11C) and CKX3/cue1-6 (Figure 11D) were also smaller compared to cue1-1 or cue1-6, indicating that active cytokinins are depleted in the shoot due to CKO activity. However, root growth was even further inhibited in CKX3/cue1-6 compared to cue1-6 (Figure 11D), suggesting that cytokinin accumulation in the roots of the mutant is not the cause for the stunted root phenotype.

In order to further characterize the root- and leaf phenotypes of cue1, the spectra of amino acids in both organs of cue1 as well as in lcd1 were analyzed (Figures 12A–H, Supplemental Tables 1A–L). There were profound differences in the amino acid composition in rosette leaves depending upon whether the plants were grown on soil (Figures 12A,D) or on ½MS agar (Figures 12B,E). The amino acid spectrum in roots was determined for plants grown on ½MS agar (Figures 12C,F). Statistical ANOVA/post-hoc analyses of the amino acid data are contained in Supplemental Table 1.

Strikingly, total contents of proteogenic amino acid in leaves and roots of cue1 were almost doubled on a fresh weight basis compared to the control plants, irrespectively whether the plants were grown on soil or agar. This increase is mainly based on elevated levels of major amino acids such as Glu, Gln, Asn, Ala, and Ser, whereas Asp or Gly were not appreciably affected. In contrast, lcd1-1 exhibited similar amino acid levels in leaves and roots as the wild-type with the exception of rosette leaves of agar-grown plants, which contained slightly higher total amino acid levels, based on an increase in Gln und Gly (Supplemental Table 1E). The amino acid spectra were not only dependent on the growth conditions, but also on the ecotype of the plants (i.e., Col-0 for cue1-6 and lcd1-1 or Bensheim [pOCA] for cue1-1 and cue1-3). Arg was the only amino acid that showed a significant increase in its absolute and relative contents in leaves of cue1-1 and cue1-6, irrespectively whether the plants were grown on soil or ½MS agar (Figures 12A,B; Supplemental Tables 1A,C,E,G). However, in roots of cue1-6 or cue1-1, Arg contents were not affected and exhibited similar or even lower levels compared to the control plants (Figures 12C,G; Supplemental Tables 1A,C,E,G). All three AAA were significantly decreased in leaves of ½MS agar-grown cue1-6 (Figures 12B,E), whereas in leaves of soil-grown cue1-6 AAA were less prominently increased (Figures 12A,D), compared to the other amino acids. The latter also applies to cue1-1, with the exception of Phe, which was slightly decreased. The amino acid spectra of soil- and agar-grown lcd1-1 delivered ambiguous results, as some amino acids such as His and Arg were decreased in soil-grown plants, but increased when the plants were grown on ½MS agar.

There were interesting changes in the amino acid composition in roots compared to leaves of cue1, in that contents of AAA or branched-chain amino acids were increased rather than decreased, or remained unaltered compared to the wild type (Figures 12C,F). Moreover, Ser contents were increased in the strong cue1 alleles cue1-6 and cue1-1 (Figure 12C; Supplemental Table 1). In contrast, roots of lcd1-1 exhibited a decline in Arg and Trp and an increase in the major amino acids Glu, Gln and Asp (Figure 12C, Supplemental Table 1).

In order to gain more information on the metabolic basis for the rescuing effect of AAA and tZ on the leaf phenotype of cue1, the amino acid composition was determined in leaves of Col-0, cue1-6 and lcd1-1, grown on ½MS agar either in the absence or presence of AAA or tZ (Figures 12G–K). In the presence of an AAA cocktail, the content of endogenous AAA was increased substantially 70- to 200-fold in wild-type and mutant plants. The three AAA summed up to 20, 23 and 18 μM·g⁻¹ FW in Col-0, cue1-6 and lcd1-1 respectively (Supplemental Table 2A) and thus led to an increase in the total amino acid content (Figure 12H; Supplemental Table 2A). Feeding of AAA resulted in a general decrease in Asn, Leu, and Ile, whereas Glu, Asp, and Ser showed a moderate increase in all plant lines (Figure 12I). There was a specific increase of Gly in cue1-6 and a drop in Arg contents in both cue1-6 and lcd1-1 when grown on AAA (Figure 12I). Despite this decrease, the Arg content was still higher in cue1 compared to the wild type (Figure 12H).

Strikingly, growth on tZ resulted in an increase in total amino acid contents in both Col-0 and lcd1-1, but a moderate decrease in cue1-6 (Figure 12K). Moreover, there was a similar pattern in tZ dependent changes in the amino acid composition in cue1-6 and lcd1-1. Despite the rescue of the reticulate leaf phenotype in cue1, feeding of tZ resulted in a decrease of the AAA Phe and Tyr, as well as all three branched-chain amino acids. Furthermore, His and Arg were also decreased in both mutants, whereas Gly and Trp were increased.

The most surprising finding of the grafting experiments was the lack of any obvious communication between roots and shoots with reference to the individual phenotypes of both organs in cue1 (Figure 2). If long distance signaling were involved in the aberrant development of the mesophyll or the stunted roots, an at least partial rescue of the individual cue1 phenotypes would have been expected, depending upon whether cue1 scions were grafted on wild-type root stocks or vice versa. DCG has been proposed earlier as long distance signaling molecule (Voll et al., 2003). Indeed, the application of DCG rescued the reticulate leaf phenotype of cue1 (due to the lack of availability, the effect of DCG on lcd1-1 leaves has not been analyzed). However, like in the tobacco mesophyll cell culture system (Tamagnone et al., 1998a,b), DCG appeared to exert only a local effect in cue1 leaves. It remains to be elucidated whether A. thaliana contains detectable amounts of DCG. In contrast to the tobacco system, DCA (the aglycon of DCG) could not rescue the cue1 phenotype, which suggests that the appropriate glucosyl transferase was either not active or not present in A. thaliana. Moreover, like for DCA the non-glycosylated forms of two more lignans, pinoresinol, or lariciresinol, also failed to rescue the reticulate leaf phenotype. More substances ought to be tested to either support or rule out an involvement of lignans in the impaired mesophyll development of cue1. It has been shown earlier, that DCG can exert cytokinin-like effects in tobacco callus cultures (Teutonico et al., 1991). Therefore, we tested the effect of cytokinin feeding on the phenotype of cue1 and lcd1.

Cytokinin metabolism is complex and comprises, besides of highly active compounds like tZ, isopentenyladenine (iP) and perhaps cZ, less active or inactive metabolites like ribosides, nucleotides and glucosides (Mok and Mok, 2001). The highly active cytokinin tZ was capable of rescuing the reticulate leaf phenotype of cue1, but not of lcd1 (Figure 6). Unlike AAA or DCG, the latter of which eventually derives from Phe as one of the end products of the shikimate pathway, cytokinins are based on purine biosynthesis (Mok and Mok, 2001). Thus, in contrast to AAA, an obvious link of tZ to a perturbed plastidial PEP metabolism in cue1 is missing. Although there are indications for a crosstalk between cytokinins and phenylpropanoids in different plant species, such as tobacco (Gális et al., 2002), maize (Alvarez et al., 2008), or A. thaliana (Bhargava et al., 2013), the exact nature of this crosstalk has not yet been resolved. It is, for instance not, clear whether cytokinins act upstream of AAA or their follow-up products (e.g., by inducing biosynthetic reactions of phenylpropanoid metabolism) or downstream of AAA (e.g., by increasing cytokinin biosynthesis or by inhibiting its degradation or conjugation). It is a matter of speculation whether the rescuing effects of both DCG and tZ converge further downstream of hormonal signaling and gene regulation.

In contrast to cue1, the reticulate phenotype of lcd1 could neither be rescued by tZ nor by AAA. This observation suggests that there are either different developmental bases for this phenotype in A. thaliana, or the defect in lcd1 is located either down- or upstream of the point where AAA and/or cytokinins are able to rescue the leaf phenotype of cue1. It has been shown by crossing experiments, that cue1 is epistatic to lcd1 (or re-3), suggesting that both proteins are involved in the same developmental pathway (González-Bayón et al., 2006). However, a more simplistic view on the rescue of the reticulate leaf phenotype by tZ or AAA would be that the cell cycle in the mesophyll of cue1 is more readily inducible compared to lcd1 or the wild type. This assumption is supported by the observation that feeding of both AAA and tZ significantly increased the cell number in the mesophyll of cue1 (Figure 6) due to an enhanced mitotic activity (not shown). Enhanced cell numbers have neither been observed in wild-type nor in lcd1 plants. The hormonal control of the cell cycle in A. thaliana comprises an interaction of auxins, cytokinins, ABA and brassinosteroids, which either induce or repress cell cycle regulators (Gutierrez, 2009). It remains to be elucidated whether the regulation of the cell cycle is altered in cue1.

Bearing the lack of any similarities between cue1 and lcd1, with reference to the rescue by tZ, in mind, it was surprising that the pattern of changes in the amino acid composition as a response to tZ feeding was very similar in both mutants (Figure 12K). In particular branched-chain amino acids, as well as Phe, Tyr, His, and Arg were decreased, whereas Trp and Gly were increased. In particular Arg contents in leaves of cue1 were consistently elevated in our study (Supplemental Tables 1, 2) and in earlier reports (Streatfield et al., 1999), and are, hence, back to wild-type level by feeding of AAA or tZ. Increased Arg contents have been proposed to be involved in nitric oxide (NO) production in the nox1 mutant, which is allelic to cue1 (He et al., 2004). NO is an important signaling molecule in plants and can, for instance, interfere with phenylpropanoid metabolism and thereby regulate responses to pathogen attacks (Zeier et al., 2004). The synthesis of NO was proposed to be either catalyzed by NO synthase (NOS) via Arg and citrulline or by a side reaction of nitrate reductase via nitrite (Wendehenne et al., 2001; del Rio et al., 2004). A loss of function mutant of NOS (nos1) contained lower NO contents in roots (Guo et al., 2003). In another system, feeding of the NO donor sodium nitroprusside resulted in a further inhibition of root growth in nox1 (cue1), but not in the wild type (He et al., 2004). More recent findings suggest that the NOA (NITITE OXIDE ASSOCIATED PROTEIN1)/RIF1 (RESISTANT TO INHIBITION BY FOSMIDOMYCIN) protein, a plastidial GTPase, is involved in the regulation of NO synthesis (Gas et al., 2009).

In contrast to cue1, the ven3/ven6 mutants defective in CPS and, hence, Arg biosynthesis show lower Arg levels (Mollá-Morales et al., 2011). Thus modified Arg levels are not per se responsible for the reticulate leaf phenotype. In line with the defect in cue1, the reticulate trp2 mutant, impaired in Trp biosynthesis, is deficient in the synthesis of AAA. Further comparative analyses are required to reveal common mechanistic bases for the developmental constraints in the mesophyll. A detailed update on reticulate mutants is contained in Lundquist et al. (2013). Furthermore, these authors discuss a “supply” or “signaling” hypothesis (see also Rosar et al., 2012), which are both based on the assumption that the bundle sheath of C3 plants plays a profound role in mesophyll development. Although bundle sheath cells in C3 plants are morphologically similar to mesophyll cells, they can conduct a distinct metabolism (Hibberd and Quick, 2002), or their chloroplast carry transporters, like for instance GPT1 (Kunz et al., 2010) or PPT1 (Knappe et al., 2003b), which are not or only weakly expressed in mesophyll cells.

Although AAA and tZ were capable of rescuing the reticulate leaf phenotype, they further deteriorated the stunted root phenotype of cue1 and also inhibited root growth in wild-type and control plants (see Figures 7, 8). Although the endogenous tZ contents in cue1 roots were more than doubled compared to wild-type or control roots, it is less likely that elevated cytokinin levels contribute to the stunted root phenotype. For instance, overexpression of the cytokinin-degrading enzyme CKO3 in the cue1 background could not rescue the root phenotype of cue1. In contrast, root growth was even further inhibited in the overexpressing lines (Figure 11E). It has been shown earlier that overexpression of CKX1 and CKX3, which are localized in the vacuole, exerted the strongest cytokinin dependent deficiency syndrome, such as promotion of root growth (Werner et al., 2003).

The most pronounced inhibitory effect on root elongation by individual AAA was observed after feeding of Trp, which resulted in an arrest of root growth. This might be due to feedback inhibition of the initial step of the shikimate pathway by Trp (e.g., Herrmann and Weaver, 1999) and hence a depletion in Phe and Tyr in the presence of Trp. Moreover, Trp is also the precursor for auxin biosynthesis (e.g., Ljung, 2013) and thus a link between amino acid metabolism and hormonal signaling. The observation that cue1 was less responsive to the inhibitory effect of NAA on root growth might point into the direction of hormonal signaling. Although endogenous auxin levels in roots of cue1 were not different from wild-type or control plants (Table 1), an altered sensitivity toward auxin might contribute to the stunted root phenotype, an observation that deserves more attention in future experiments. Cytokinins can interfere with the expression of auxin related genes both in roots and shoots (Brenner and Schmülling, 2012). Probably increased endogenous tZ levels in roots of cue1 are involved in the decreased auxin sensitivity. Endogenous levels of Trp were more steeply increased in roots of the cue1-1 allele compared to cue1-6 or the weak cue1-3 allele. Thus, Trp accumulation in the roots correlated to some extend with the inhibition of root growth in the individual cue1 mutant alleles (compare Figures 8B, 12C). Moreover, lcd1, which contained lower levels of Trp, developed longer roots (Figures 8B, 12C). Again, any direct link between Trp contents, sensitivity toward auxin, and root length in cue1 or lcd1 remains to be established.

Plants with an impaired phospho-Ser (P-Ser) pathway of de novo Ser biosynthesis also display a stunted root phenotype (Benstein et al., 2013). Due to a lesion in the initial step of this pathway, i.e., in phosphoglycerate dehydrogenase (PGDH), particularly in PGDH1, Ser biosynthesis in roots and other non-photosynthetic tissues is hampered. As Ser is also required for Trp biosynthesis, the stunted root phenotype of PGDH1 knockdown plants probably reflects diminished Trp and concomitantly auxin levels in the roots. On a whole seedling scale, IAA levels were clearly diminished in the knockdown plants compared to the wild type (Benstein et al., 2013). However, in the case of the PGDH1 knockdown lines, the demand for Trp seems not to be the only reason for the root growth defects, as feeding with Trp did not rescue the phenotype.

The pleiotropic phenotypes of the cue1 mutant, which are manifested independently in different organs, such as leaves and roots, point at the necessity to re-evaluate the putative roles of the PPT in plastids of autotrophic and heterotrophic tissues. It is well established and supported by experimental data that most plastids, particularly chloroplasts, rely on the import of PEP not only for the shikimate pathway, but also for plastidial pyruvate-derived pathways like de novo fatty acid biosynthesis, the synthesis of Val and Leu and the MEP pathway. Strikingly, homozygous double mutant with a combined loss of function in PPT1 and ENO1 were lethal (Prabhakar et al., 2010). Likewise, the inability to produce pyruvate from PEP in PK deficient plants (Andre et al., 2007; Baud et al., 2007) resulted in diminished oil contents of the seeds. Unlike mixotrophic plastids, i.e., those in embryos during the early stages of seed development, which are able to provide PEP both by a complete plastidial glycolysis and by import from the cytosol, chloroplasts are entirely dependent on PEP import via the PPT, because chloroplasts lack ENO1 (Prabhakar et al., 2009). In contrast, ENO1 expression in roots is appreciably high and thus can support PEP provision by plastidial glycolysis. Hence, PEP import into root plastids by a PPT might not be necessary. Moreover, PEP accumulation inside the plastids due to a lack in the PPT as PEP exporter might even be harmful to the plants and results in stunted roots. There are several lines of evidence that support this idea. (1) If a depletion in products of the shikimate pathways deriving from plastidial PEP were the major cause for the stunted root phenotype of cue1, wild-type scions grafted on cue1 root stocks would eventually rescue the phenotype as, for instance, aromatic amino acid as well as branched-chain amino acids can be transported via the phloem (Winter et al., 1992). (2) The absolute contents of AAA as well as of the branched-chain amino acid Leu are increased in cue1 roots as PEP can be generated via glycolysis, e.g., from imported Glc6P (Kunz et al., 2010). (3) Ser as product of the P-Ser pathway (Benstein et al., 2013; Cascales-Miñana et al., 2013) is increased in cue1 roots, probably because products of plastidial glycolysis downstream of 3-PGA, the substrate of the P-Ser pathway, accumulate. (4) Increased Ser contents correlate with increased levels of Trp, which has been shown to be toxic to root growth at high concentration. (5) The prenyl side chain of the cytokinin tZ, which is substantially increased in cue1 roots, derives from the plastidial MEP pathway (Kasahara et al., 2004).

Figure 13 shows the proposed consequences a deficiency in the PPT1 has on leaf and root metabolism. In chloroplasts, the lack of PPT1 can be partially compensated by PPT2 (Figure 13B), which is highly expressed in leaves, but not in roots. The analysis of double mutants defective in both PPT1 and PPT2 might shed further light on the organ specific and developmental roles of both transporters. As the PPT acts as an importer in chloroplasts (Figure 13A), products deriving from either plastidial PEP, such as AAA, or pyruvate, like fatty acids, branched-chain amino acids and products of the MEP pathway are diminished (indicated by the green background color). However, it has been shown recently that the MEP pathway is hampered in plants with an impaired Na-pyruvate transporter (Furumoto et al., 2011) suggesting that pyruvate as one of the substrates of the MEP pathway can be imported directly from the cytosol. Hence, products deriving from plastidial pyruvate are shown with a light green background color (Figure 13B). Due to the presence of ENO1 in root plastids (Figure 13C), PEP import from the cytosol is not required. Moreover, in contrast to chloroplasts, a knockout of PPT1 cannot be compensated by PPT2 (Figure 13D). We propose that the PPT in root plastids acts as an overflow valve for PEP and/or 2-PGA and thus connects plastidial with cytosolic glycolysis. In the absence of the PPT, products deriving from PEP and pyruvate metabolism accumulate (indicated by a red background colors). Key enzymes of the MEP pathway, 1-deoxy-D-xylulose 5-phosphate reductoisomerase or 1-deoxyxylulose 5-phosphate synthase show a substantial expression in roots (Estévez et al., 2000; Carretero-Paulet et al., 2002; eFP browser, Winter et al., 2002). Likewise, fatty acid biosynthesis and the production of Val and Leu are increased. The branched-chain amino acid Ile is probably less affected, as it is synthesized from Thr and eventually from Glu, which in root plastids is provided by NADH-dependent glutamate synthase. The redox equivalents required for this reaction are provided by a plastidial NAD malate dehyrogenase (Selinski et al., 2014).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.