The Arogenate Dehydratase ADT2 is Essential for Seed Development in Arabidopsis

Abstract

Phenylalanine (Phe) biosynthesis in plants is a key process, as Phe serves as a precursor of proteins and phenylpropanoids. The prephenate pathway connects chorismate, the final product of the shikimate pathway, with the biosynthesis of Phe and tyrosine. Two alternative routes of Phe biosynthesis have been reported: one depending on arogenate, and the other on phenylpyruvate. Whereas the arogenate pathway is considered the main route, the role of the phenylpyruvate pathway remains unclear. Here, we report that a deficiency in ADT2, a bifunctional arogenate dehydratase (ADT)/prephenate dehydratase (PDT) enzyme, causes embryo arrest and seed abortion. This result makes a clear distinction between the essential role of ADT2 and the five remaining ADT genes from Arabidopsis, which display mostly overlapping functions. We have found that PHA2, a monofunctional PDT from yeast, restores the adt2 phenotype when it is targeted within the plastids, but not when is expressed in the cytosol. Similar results can be obtained by expressing ADT3, a monofunctional ADT. These results suggest that Phe can be synthesized from phenylpyruvate or arogenate when the bifunctional ADT2 is replaced by other ADT or PDT enzymes during seed formation, highlighting the importance of Phe biosynthesis for embryo development, and providing further insights into the plasticity of Phe biosynthesis.

Introduction

The biosynthesis of phenylalanine (Phe) is an essential process in plant metabolism and physiology due to its critical function as a building block for proteins but also as the main precursor of phenylpropanoids, a wide range of aromatic compounds essential for plant growth, development and stress responses (Maeda and Dudareva 2012). It is well accepted that the emergence of specialized metabolic pathways, such as phenylpropanoid metabolism, was fundamental for land colonization by the first terrestrial plants, a critical step in the history of life (Emiliani et al. 2009, Weng and Chapple 2010, Renault et al. 2017)

Within plant plastids, the prephenate pathway connects chorismate, the final product of the shikimate pathway, with the biosynthesis of the aromatic amino acids Phe and tyrosine (Tyr) (Maeda and Dudareva 2012) (Fig. 1). In particular, the biosynthesis of Phe can take place through two alternative branches of the prephenate pathway: the arogenate pathway, that is considered to be the common in land plants, and the phenylpyruvate pathway, which is characteristic of most microorganisms. In the arogenate pathway, prephenate is transaminated by prephenate-aminotransferase (PAT) to generate arogenate, which is decarboxylated and dehydrated by arogenate dehydratase (ADT) to yield Phe (Bonner and Jensen 1987a). In the phenylpyruvate pathway, the decarboxylation and dehydratation of prephenate generates phenylpyruvate under the action of prephenate dehydratase (PDT). Phenylpyruvate is subsequently transaminated to Phe through a phenylpyruvate aminotransferase (Fischer and Jensen 1987a). Prephenate and arogenate are also precursors for the biosynthesis of Tyr through the alternative action of either prephenate dehydrogenase (PDH) or arogenate dehydrogenase (ADH) enzymes (Fig. 1; Bonner and Jensen 1987b, Fischer and Jensen 1987b).

Fig. 1

The prephenate pathway in plants. The arogenate branch (shadowed) is considered to be the major source of Phe in plants. Enzyme abbreviations: CM, chorismate mutase; PDH, prephenate dehydrogenase; PDT, prephenate dehydratase; PAT, prephenate aminotransferase; AAA, aromatic amino acid amino transferase; ADH, arogenate dehydrogenase; ADT, arogenate dehydratase.

Despite the biochemical and genetic evidence that indicate that the arogenate pathway is the primary route for Phe biosynthesis in plants (Maeda et al. 2010, Maeda et al. 2011), various reports had suggested in the last few years that PDT activity may also contribute to the synthesis of this amino acid in plants under certain conditions (Warpeha et al. 2006, Yoo et al. 2013, Oliva et al. 2017). In this regard, the molecular and kinetic characterization of the ADT family from Arabidopsis thaliana (Cho et al. 2007) and, more recently, Pinus pinaster (El-Azaz et al. 2016), has demonstrated that some ADT isoforms can also display PDT activity in vitro, at lower catalytic efficiencies compared with the alternative use of arogenate as substrate. In addition, these ADT/PDT bifunctional proteins can rescue the Phe auxotrophic phenotype caused by the suppression of the endogenous PDT activity in the yeast mutant pha2 (Bross et al. 2011, El-Azaz et al. 2016), strongly supporting that these ADT proteins can work as PDTs in vivo. Further evidence about the existence of a functional PDT pathway in plants is provided by the accumulation, in response to the provision of exogenous shikimate, of phenylpyruvate and the restoration of the Phe levels in Petunia hybrida petals silenced for ADT activity (Maeda et al. 2010). Moreover, Yoo et al. (2013) described a cytosolic phenylpyruvate-aminotransferase in P. hybrida that efficiently catalyzes the interconversion of phenylpyruvate and Tyr to Phe and 4-hydroxyphenylpyruvate, respectively. Thus, these authors proposed the putative existence of a functional phenylpyruvate pathway in plants and provided evidence of its functionality when plants were transgenically manipulated to block the flux through the arogenate pathway (Yoo et al. 2013). Consistently, in a previous report, we showed that silencing of the arogenate pathway in Nicotiana benthamiana leaves, through virus-induced gene silencing of PAT, resulted in the overexpression of two ADT/PDT genes coding for putative ADT/PDT enzymes (de la Torre et al. 2014).

However, despite the evidence that has been provided regarding the existence of a functional phenylpyruvate pathway in plants, we lack direct genetic evidence about the participation of PDT activity leading to Phe biosynthesis in Arabidopsis. Here we report an embryo-arresting and seed abortion phenotype caused by the genetic disruption of the ADT2 gene (At3g07630) in A. thaliana, that encodes a bifunctional ADT/PDT protein, demonstrating that this gene is essential for completing the life cycle of the plant. This result makes a clear distinction between ADT2 and the five remaining ADTs from A. thaliana, which had been reported to have partially overlapping roles in different aspects related to plant growth and development such as lignin accumulation or the size and mass of the stems (Corea et al. 2012a, Corea et al. 2012b). Very interestingly, the authors also carried out a co-expression analysis of ADT genes and determined that ADT3, ADT4, ADT5 and ADT6 are co-regulated with genes involved in the shikimate pathway and the biosynthesis of aromatic amino acids and phenylpropanoids, while ADT1 and ADT2 are predominantly co-expressed with genes involved in basic cellular functions such as transcription or translation (Corea et al. 2012b). In the present work, we have found that the Saccharomyces cerevisiae monofunctional PDT enzyme PHA2 (Bross et al. 2011), and ADT3, a monofunctional ADT from Arabidopsis (Cho et al. 2007), can rescue the adt2 mutant phenotype when expressed under the control of the putative ADT2 promoter region and targeted to plant plastids, but not when expressed in the cytosol. The functional complementation of the adt2 mutant by a PDT enzyme provides genetic evidence about the contribution of the phenylpyruvate pathway to Phe biosynthesis, and supports the compartmentalization model proposed by Yoo et al. (2013). Taken together, our results demonstrate that the single reintroduction of either of the two enzymatic activities of ADT2, ADT or PDT, allows rescue of the adt2 lethality phenotype, suggesting that the biosynthesis of Phe during Arabidopsis seed formation could be carried out from both arogenate and phenylpyruvate under the action of ADT2. These findings highlight the importance of Phe metabolism in plant reproduction, either for protein or phenolic biosynthesis, and its compartmentalization at the subcellular level, and emphasize the potential contribution of the phenylpyruvate pathway to Phe production in land plants.

Results

ADT2 is an essential gene for seed formation

Previous reports had demonstrated the possibility of generating homozygous T-DNA knock-out mutants for all the Arabidopsis ADT genes and their different combinations, except ADT2, revealing a partially redundant role for all members except ADT2 in plant growth and development (Corea et al. 2012a, Corea et al. 2012b, Chen et al. 2016). The inability to generate plants knocked out for ADT2 clearly points to an essential role for this protein during the Arabidopsis life cycle.

To investigate further in what way ADT2 can be essential for the plant, we obtained from the Arabidopsis Biological Resource Center (ABRC) two alternative T-DNA insertion mutation lines for this gene: SALKseq_044042 (adt2-1, insert position at exon number 6) and SALKseq_081342 [adt2-2, insert position at –34 bp from the open reading frame (ORF) start codon] (Supplementary Fig. S1). In parallel, we also obtained T-DNA lines for the remaining ADT genes, ADT1 and ADT3–ADT6 (Supplementary Table S1). Insertion sites for each mutant line were located by using the information available in the SIGnAL database (http://signal.salk.edu/cgi-bin/tdnaexpress) and all lines were verified by PCR (Supplementary Fig. S1). Regarding the adt2-1 and adt2-2 lines, the heterozygosity was confirmed for both lines (Supplementary Fig. S1).

To elucidate further the heterozygosity observed for the adt2 mutant lines, 215 seedlings for the adt2-1 line and 250 for the adt2-2 line, originating from independent parental heterozygous plants, were analyzed for their respective T-DNA insertion in the ADT2 gene (Supplementary Fig. S2). No homozygous adt2-1 or adt2-2 mutants were detected in this screening, strongly suggesting that the mutant alleles adt2-1 and adt2-2 cannot be fully transmitted through one of the gamethophytes or cause seed lethality. This result was confirmed repeatedly after several generations.

Heterozygous adt2 plants produced shorter siliques with a reduced seed set compared with the wild type (Fig. 2a; Supplementary Fig. S3). Although these plants start to develop normal seeds after pollination, about 25% of the seeds turn yellowish and comparatively more transparent as silique development continues, finally turning brown and shriveled after a few days (Fig. 2b). The rate of aborted seeds was counted in the adt2-1 (n = 1,109) and adt2-2 (n = 1,232) siliques, giving an adjusted ratio of 3:1 (viable seeds:aborted seeds; Fig. 2c), which is consistent with mutations that affect zygotic development (Page and Grossniklaus 2002). Interestingly, this phenotype appeared to be characteristic of ADT2 deficiency, since it was not observed in any of the adt1 and adt3–adt6 mutant lines. We were not able to analyze kanamycin resistance segregation in the adt2-1 and adt2-2 progeny, since the resistance marker seems to be silenced in the adt2-1 line, whereas all the plants in the adt2-2 line were resistant albeit PCR demonstrated heterozygosity in the adt2 insertion, revealing the existence of a second, independent T-DNA insertion in an unknown location. Additionally, reciprocal crosses between wild-type plants and heterozygous adt2-1 or adt2-2 mutants resulted in a segregating progeny with heterozygous adt2 descendants, confirming that the adt2 mutation can be equally transmitted by male or female gametes. Our phenotypic characterization of ADT2 mutant plants also showed that heterozygous adt2 plants have a prostrate phenotype which has been previously observed in other ADT mutants (Corea et al. 2012b; Supplementary Fig. S2).

Fig. 2

Phenotypic characterization of adt2 mutant plants. (a) Mature siliques from the different ADT gene family T-DNA insertion mutants, showing the existence of an abnormal phenotype in the siliques from the adt2 mutants analyzed, which is absent in the other ADTs. Pictures were taken at the same magnification. (b) Seed phenotype in the adt2 heterozygous mutants. Yellowish seeds (top) become brown and shriveled when siliques start drying (bottom). (c) Seed abortion rate in the adt2 heterozygous mutant lines. χ²-test confirmed the 3:1 segregation (normal seeds:aborted seeds) of the abortion phenotype. WT: wild-type plants.

adt2 mutants exhibit embryo arrest at the globular stage

The analysis of the adt2 progeny strongly suggested that the ADT2 deficiency causes lethality, since it was not possible to obtain homozygous adt2 mutants from either adt2-1 or adt2-2 lines (Supplementary Fig. S2). The seed abortion rate in the adt2 plants, that matches an adjusted 3:1 ratio (viable seeds:aborted seeds; Fig. 2c), indicates that this lethality could be related to zygote formation or embryo development. To reinforce these analyses, seeds from adt2 heterozygous plants were analyzed through differential interference contrast microscopy (DIC) (Fig. 3). In young siliques, all the seeds contained normal embryos at the early globular stage (Fig. 3a,d). Nevertheless, whereas most of the embryos continued to develop normally to the heart stage (Fig. 3b), retarded embryos started to occur in certain positions (Fig. 3e). The embryo arrest phenotype becomes evident in older siliques, when wild-type-appearing seeds contain well-formed mature embryos (Fig. 3c) and mutant-type seeds contain embryos arrested at the globular stage (Fig. 3f,g).

Fig. 3

Embryos from the adt2 heterozygous mutants are arrested at the globular stage. (a–c) Embryos of wild-type-like seeds of the adt2 heterozygous mutant, in the early globular stage (a), heart stage (b) and cotyledon stage (c). (d–f) Non-viable seeds from the same respective siliques as (a), (b) and (c), showing embryos arrested at the globular stage. (g) Detail from an arrested embryo in the same developmental stage as in (e). Scale bars = 50 µm.

Previous reports have linked early embryo-arresting phenotypes to deficiencies in female gametophyte formation (Pagnussat et al. 2005, Rabiger and Drews 2013), providing evidence that gene expression from the female gametophyte or the maternal genome can play a key role in early embryogenesis. To test this possibility, we managed to isolate, through micro-manipulation techniques, ovules from adt2 plants in different developmental stages (Supplementary Fig. S4). Female gametophytes from adt2 mutants exhibited a normal phenotype, suggesting that adt2 mutation is not disturbing the correct patterning of the maternal organs, and therefore arguing against the maternal background of the adt2 embryonic arrest.

Expression pattern of ADT2

Based on the data available from the Arabidopsis expression atlas (Arabidopsis eFP Browser 2.0; Winter et al. 2007) ADT2 is nearly ubiquitously expressed in the whole plant at diverse stages of its life cycle (Supplementary Fig. S5), showing the highest expression levels in dry seeds and imbibed seeds and, to a lesser degree, in the later stages of seed development (stages 8–10), shoot apex inflorescences and senescent leaves. In the embryo, ADT1 and ADT2 are the most abundant ADT isoforms (Supplementary Fig. S5). Interestingly, the data available indicate that ADT2 expression peaks in the transition from the globular stage into the heart stage of the embryo, more particularly in the basal region of the embryo. This patterning of ADT2 expression might be associated with the arresting phenotype observed at this stage.

To corroborate this information about ADT2 expression during seed development, we performed histochemical studies using the β-glucuronidase (GUS) gene under control of a 471 bp fragment containing the putative 5’-upstream regulatory region of ADT2 (P_ADT2). This region was subcloned into the pGWB3 vector and the resulting P_ADT2-GUS construct was used to transform Arabidopsis Col-0 plants. Analysis of three independent P_ADT2-GUS reporter lines showed similar results (Fig. 4). In contrast to the in silico analysis of ADT2 expression in the embryo, P_ADT2-GUS reporter lines did not exhibit a peak in the GUS staining level in the transition from the globular stage into the heart stage of the embryo (Fig. 4b,c). GUS staining indicates that P_ADT2-driven expression is, instead, mostly spatially and temporally uniform during the main stages of embryo development. In addition, GUS signal was also found to be present in other seed structures, such as the inner integument of the seed coat (Fig. 4f). GUS staining of tissues corresponding to transgenic lines carrying the control pGWB3 empty vector showed a very slight and non-specific background signal (Supplementary Fig. S6).

Fig. 4

GUS staining of transgenic P_ADT2-GUS plants revealing the ADT2 expression pattern during seed development. Embryos at (a) early globular, (b) late globular, (c) heart, (d) torpedo and (e) cotyledon stages. (f) Detail of the seed coat layers during the transition from the late globular to the heart stage. oi, outer integument; ii, inner integument. Scale bars = 50 µm.

Complementation of the adt2 phenotype by ADT2

Our analysis of the adt2-1 and adt2-2 mutant lines revealed that ADT2 could be essential for correct seed formation. To ascertain whether the adt2 arrested embryos are the result of a deficiency in ADT2, heterozygous adt2-1 mutants were transformed with the plant expression vector pGWB610 (no promoter, C-terminal FLAG tag; Nakamura et al. 2010) containing the ADT2 full-length cDNA under the control of P_ADT2. Transgenic pGWB610::P_ADT2-ADT2 lines (from hereon, P_ADT2-ADT2) were selected for BASTA resistance, and T₂ plants from three independent transgenic lines were analyzed by PCR, showing the existence of homozygous adt2 individuals. T₃ progeny from the T₂ putatively homozygous adt2 plants were grown and analyzed by PCR, confirming the homozygous adt2 genotype of their respective parental plants (Fig. 5a). In addition, homozygous adt2 mutant plants were analyzed by quantitative reverse transcription–PCR (RT–qPCR), demonstrating that the endogenous ADT2 transcript is not detected in these plants (Supplementary Fig. S7). No homozygous adt2 mutants were found among the T₂ and T₃ progeny of control plants transformed with the empty vector. Moreover, T₃ plants transformed with the P_ADT2-ADT2 construct showed siliques without aborted seeds (Fig. 5c,d), demonstrating the ability of P_ADT2-ADT2 to rescue the adt2 mutant phenotype. P_ADT2-ADT2 expression was confirmed by Western blot for the FLAG-tagged recombinant protein (Fig. 5b).

Fig. 5

Complementation of the adt2 mutant phenotype. (a) PCR test of T₃ adult plants, confirming the existence of homozygous adt2-1 mutants when transformed with ADT2, PHA2 or ADT3 under control of the ADT2 promoter (P_ADT2). Numbers from 1 to 3 indicate the three independent transgenic lines selected for each construct. (b) Anti-FLAG Western blot of total protein extracts from three independent T₃ transgenic lines (1–3) expressing ADT2, ADT3 or PHA2 under control of P_ADT2. T₃ lines transformed with the empty vector were used as negative control. The predicted masses for the recombinant proteins were: 42.1 kDa (ADT2), 38.2 kDa (PHA2) and 46.1 kDa (ADT3). (c) Mature siliques from wild-type plants (Col-0) compared with T₃ T-DNA mutant plants transformed with the empty vector (E.V.), ADT2, PHA2 or ADT3 expression constructs. (d) Seed abortion rate in adt2 null mutants complemented with ADT2, PHA2 or ADT3 under control of the ADT2 promoter, compared with wild-type (Col-0) and adt2 plants transformed with the empty vector (E.V.).

Expression of PHA2 and ADT3 rescues the adt2 mutant phenotype

Although the arogenate pathway has been proposed as the main route leading to Phe biosynthesis in plants, the biosynthesis of Phe from phenylpyruvate through PDT activity has been reported to occur under particular conditions (Warpeha et al. 2006, Yoo et al. 2013). Since A. thaliana ADT2 encodes a bifunctional ADT/PDT enzyme (Cho et al. 2007, Bross et al. 2011), we were interested in verifying whether the adt2 mutant phenotype could be rescued through the heterologous expression of monofunctional PDT or ADT enzymes. To test if adt2 plants can be complemented by a PDT enzyme, the S. cerevisiae PHA2 gene was expressed in heterozygous adt2-1 plants. Saccharomyces cerevisiae uses phenylpyruvate as the intermediate for the biosynthesis of Phe, and no PAT or ADT activities have been detected in this organism (Gottardi et al. 2017). In addition, PHA2 and plant ADTs/PDTs strongly differ in their primary sequence, and have been proposed to have a different evolutionary origin (Dornfeld et al. 2014), but no direct experimental evidence has been provided excluding the use of arogenate as a substrate by PHA2. For this reason, we tested in vitro the ADT activity of recombinant PHA2, demonstrating that this enzyme is not able to act as an ADT (Supplementary Fig. S8). Moreover, we determined the kinetic properties of PHA2 as a PDT, showing that recombinant PHA2 works efficiently as a PDT (K_{m prephenate} = 35 ± 5 µM; V_max = 1.261 ± 171 pKat µg^–1 of recombinant protein; Supplementary Table S2).

In order to target yeast PHA2 into Arabidopsis plastids, where Phe biosynthesis takes place (Rippert et al. 2009), a chimeric PHA2 gene was made by introducing the DNA sequence coding for the putative plastid transit peptide of ADT2 into the 5’-flanking region of the PHA2 full-length ORF. This chimeric PHA2 was cloned into the pGWB5 expression vector [Cauliflower mosaic virus (CaMV) P35S promoter; C-terminal green fluorescent protein (GFP) fusion] and the resulting construct, pGWB5::PHA2-GFP, was transiently expressed in N. benthamiana leaves via agroinfiltration. The plastidial localization of chimeric PHA2–GFP was successfully confirmed by confocal microscopy (Supplementary Fig. S9). To confirm further that heterologous PHA2 was effectively targeted into Arabidopsis plastids, we performed a stable Arabidopsis transformation with the pGWB5::PHA2-GFP construct, and determined the plastidial localization of the PHA2–GFP fusion protein (Supplementary Fig. S9).

Similarly to P_ADT2-ADT2, heterozygous adt2-1 plants were transformed with the construct pGWB610::P_ADT2-PHA2, containing the chimeric PHA2 gene described above, under the control of the P_ADT2 regulatory region. For complementation analysis, three independent P_ADT2-PHA2 transgenic lines were selected. As described for the P_ADT2-ADT2 transgenic plants, the presence of adt2-1 homozygous plants was successfully detected among the T₂ progeny of the P_ADT2-PHA2 lines. Their respective T₃ progeny confirmed the homozygous adt2-1 genotype and also showed siliques without aborted seeds (Fig. 5a–d). RT–qPCR analysis of these plants did not detect ADT2 expression, confirming adt2 homozygous mutants as null (Supplementary Fig. S7). These results indicate that the PHA2-mediated introduction of PDT activity into the adt2 mutant background, in which the PDT activity of ADT2 has been suppressed, is sufficient to rescue the adt2 mutant phenotype through a Phe biosynthetic pathway that uses phenylpyruvate as intermediate.

Furthermore, heterozygous adt2-1 plants were transformed with a pGWB610::P_ADT2-ADT3 construct, containing the complete ORF of the Arabidopsis ADT3 gene, which encodes an ADT enzyme without detectable PDT activity (Cho et al. 2007, Bross et al. 2011). The expression of ADT3 under the control of the ADT2 promoter, P_ADT2-ADT3, also resulted in the complementation of the adt2 zygotic lethality phenotype, indicating that the ADT activity encoded by ADT2 can be restored by ADT3 in the adt2 null background (Fig. 5a–d; Supplementary Fig. S7). Together, our complementation studies using P_ADT2-PHA2 and P_ADT2-ADT3 transgenic plants indicate that both individual PDT and ADT activities can recover the essential role displayed by the bifunctional ADT2 enzyme.

Recently, Bross et al. (2017) described through transient expression in N. benthamiana leaves that ADT2, but none of the other ADTs, localizes in a ring around the equatorial plane of chloroplasts. Based on this result, the authors proposed that ADT2 has a moonlighting function as a participant in the chloroplast division machinery (Bross et al. 2017). In the same work, aberrant chloroplast morphology was linked to a particular ADT2 point mutation, suggesting that this moonlighting function of ADT2 could be essential for the plant to survive. To test this hypothesis, we studied the chloroplast morphology in our homozygous adt2 knock-out plants complemented with ADT2, ADT3 and PHA2. Chloroplasts corresponding to adt2 plants complemented by the expression of ADT2 were analyzed through confocal microscopy and, as could be expected, their morphology was similar to that observed in wild-type plants (Supplementary Fig. S10). Similarly, expression in the mutant adt2 background of ADT3 or PHA2, where ADT2 is completely absent (neither endogenous nor heterologously expressed protein), also resulted in chloroplasts with wild-type morphology (Supplementary Fig. S10), suggesting that the essential role of ADT2 is not directly linked to chloroplast division.

ADT2, PHA2 or ADT3 expression in the cytosol does not rescue the adt2 mutant phenotype

Intriguingly, it is well established that some enzymes of the shikimate pathway have cytosolic isoforms in plants, despite the fact that it is widely accepted that the full set of enzymes required for the biosynthesis of aromatic amino acids can only be found within the plastids (Rippert et al. 2009, Maeda and Dudareva 2012). Flowering plants possess a cytosolic form of chorismate mutase (Fig. 1; Benesova and Bode 1992, Eberhard et al. 1996, Kroll et al. 2017), CM2, whose exact role in plant metabolism remains elusive. We considered that adt2 lethality could provide a good genetic tool to test in planta the hypothetical participation of CM2 in a cytosolic pathway leading to Phe production, as its transcript has been previously detected in the embryo (Arabidopsis eFP Browser 2.0, Winter et al. 2007, Palovaara et al. 2017). For this reason, we decided to express, in the heterozygous adt2 background, the ADT2, ADT3 and PHA2 genes targeted to the cytosol (cytADT2, cytADT3 and cytPHA2; Supplementary Fig. S9). P_ADT2 or CaMV P35S promoters were separately used for each one of the three genes tested, in order to determine if the expression level of the cytosol-targeted proteins could act as a constraining factor. The occurrence of adt2 null mutants without evidence of seed abortion was checked in a similar way to that described for the plastid-targeted constructs. For each of the six cytosolic constructs (P_ADT2-cytPHA2, CaMV P35S-cytPHA2, P_ADT2-cytADT2, CaMV P35S-cytADT2, P_ADT2-cytADT3 and CaMV P35S-cytADT3), 50 individual plants from the T₂ progeny of three independent transgenic lines were analyzed by PCR to detect homozygous adt2 mutants. Neither of the cytosol-targeted enzymes, regardless of the promoter used, was able to rescue the adt2 phenotype, indicating that the introduction of cytosolic ADT or PDT activities has no functional impact in this case.

Discussion

Terrestrial plants are able to synthesize Phe through two alternative branches of the prephenate pathway, with a differential use of arogenate or phenylpyruvate as biosynthetic intermediates (Fig. 1). Most biochemical and genetic evidence indicates that the arogenate pathway is the primary route for Phe biosynthesis in plants (Maeda et al. 2010, Maeda et al. 2011). Nevertheless, different authors have suggested that, under particular conditions, the phenylpyruvate branch could contribute to the synthesis of Phe in a significant way (Warpeha et al. 2006, Yoo et al. 2013, de la Torre et al. 2014, Oliva et al. 2017). However, despite some evidence regarding the existence of a functional phenylpyruvate pathway in plants, direct genetic evidence for the participation of phenylpyruvate leading to Phe biosynthesis in Arabidopsis is lacking.

Our results indicate that ADT2 deficiency in A. thaliana causes a defect in the embryo transition from the globular stage into the heart stage, that results in seed abortion, pointing to ADT2 as essential for completing the plant life cycle. Previous reports have shown that all six Arabidopsis ADT genes are expressed nearly ubiquitously in multiple tissues with differential intensities (Cho et al. 2007, Rippert et al. 2009) and partially overlapping functions in plant development, more particularly lignin metabolism in stems (Corea et al. 2012a, Corea et al. 2012b) and anthocyanin biosynthesis (Chen et al. 2016). Interestingly, Chen et al. (2016) reported that ADT2 contributes predominantly to sucrose-induced anthocyanin biosynthesis over the five remaining ADT isoforms, with a reduction of around 65% of wild-type anthocyanin levels when ADT2 is down-regulated by artificial microRNA interference. In addition, ADT3 has been indicated as essential in etiolated seedlings for the biosynthesis of Phe that is channeled towards the production of UV-protectant pigments in response to UV irradiation (Warpeha et al. 2008). On the other hand, our results show that homozygous adt3 plants are viable and they do not seem to exhibit symptoms of abiotic stress under normal growth conditions, suggesting that its role in protection against light is not critical under such circumstances. Yet, the essential role reported in the present work for the Arabidopsis ADT2 constitutes a singular contribution that clearly differentiates this gene from the five remaining ADT genes in the plant, although other roles cannot be ruled out.

Nuclear-encoded enzymes related to plastidial metabolism of amino acids have been reported to be frequently associated with embryo lethality in Arabidopsis, which is thought to be caused by a disruption of the biosynthetic pathway(s) for such essential compounds (Bryant et al. 2011). Genes encoding enzymes related to aromatic amino acid biosynthesis are not an exception, as shown by the shikimate pathway’s bi-functional dehydroquinate dehydratase/shikimate dehydrogenase (At3g06350; EMBRYO DEFECTIVE 3004, Bryant et al. 2011; MATERNAL EMBRYO ARREST 32, Pagnussat et al. 2005) and chorismate synthase (At1g48850; EMBRYO DEFECTIVE 1144, Bryant et al. 2011) enzymes, and the prephenate pathway’s PAT (Fig. 1) (At2g22250; MATERNAL EMBRYO ARREST 22, Pagnussat et al. 2005). These enzymes have the common feature of being encoded by single-copy genes, and thus their function is hardly replaceable. Here, we have demonstrated that adt2 mutation results in embryo arrest at the globular stage (Fig. 3), incorporating ADT2 into this list of essential genes, but with the particularity that ADT2 belongs to a family with six isoforms that, in this case, do not seem to have a redundant role. Although unlikely, a minor contribution, insufficient to complement the adt2 deficiency, from other ADTs to this essential function of ADT2 cannot be ruled out.

Public data available from the Arabidopsis expression atlas (Arabidopsis eFP Browser 2.0, Winter et al. 2007) show that ADT2 is the predominant isoform detected in the transition between the globular stage and the heart stage of the embryo, more particularly in the basal region that precedes the formation of the radicle (Supplementary Fig. S5). Interestingly, our results reveal that adt2 embryos are arrested in the globular stage, suggesting that ADT2 could play a critical role in the transition from the globular to the heart stage. Nevertheless, we did not observe such a GUS staining pattern in the P_ADT2-GUS transgenic embryos (Fig. 4). Recently, a highly detailed gene expression atlas of the early Arabidopsis embryo has been published, discriminating between different cell types of the early and late globular stages of the embryo (Palovaara et al. 2017). This publication reveals that ADT2 is expressed nearly ubiquitously in the analyzed cell types corresponding to the basal region of the globular embryo (Supplementary Fig. S11), which would be in good agreement with our results, and in stark contrast to the information available from the Arabidopsis eFP browser 2.0 expression atlas. Hence, it is difficult to correlate the ADT2 expression profile in the embryo with its essential role, and to ascertain why other ADTs co-expressed in the embryo (Supplementary Figs. S4, S10) do not share such an essential role.

It is broadly accepted that developing seeds are important sinks of nutrients, including nitrogen, that are mainly incorporated into the seed as amino acids (Müller et al. 2015, Santiago and Tegeder 2016). Although all amino acids can be transported in the phloem, Glu, Gln, Asp, Asn, Ser and Leu are considered to be the predominant amino acids (Sanders et al. 2009). Once in the seed coat, these amino acids are used to synthesize the remaining proteinogenic amino acids. Finally, they are distributed towards the seed apoplast through the seed endosperm, from where they can be taken up by the growing embryo. Hence, the existing model considers that embryo amino acid provision, and in a wider sense its organic nitrogen metabolism, relies mostly on amino acid uptake from the surrounding seed tissues, rather than in situ biosynthesis in the embryo (Sanders et al. 2009, Müller et al. 2015). Given the fact that PAT and ADT2 genes seem to be essential for embryo formation, it raises the question of the regulation of the biosynthesis and transport of Phe within the seed as a whole, and the likely interplay between the embryo and the different seed tissues. Our P_ADT2-GUS assay also shows that ADT2 is expressed in the seed coat and the endosperm (Fig. 4), not discarding the fact that adt2 lethality could be related to the processes that take place in this tissues.

As mentioned before, it has been suggested that the microbial-like phenylpyruvate pathway can occur in plants to produce Phe. Yet, in petunia petals, the flux through the phenylpyruvate pathway is increased when the arogenate pathway is blocked (Yoo et al. 2013). These authors proposed a new model for the contribution of the phenylpyruvate pathway to the biosynthesis of Phe. In this model, the phenylpyruvate is synthesized within the plastid and subsequently exported to the cytosol, where the cytosolic tyrosine aminotransferase PhPPY-AT, an ortholog of AtTAT1 (At5g53970), catalyzes the conversion of phenylpyruvate to Phe in co-ordination with Tyr catabolism. Similarly, de la Torre et al. (2014) reported that N. benthamiana NbPDT1 and NbPDT2 are clearly overexpressed in tobacco leaves silenced for NbPAT. Moreover, in these plants, NbPPAT, putatively orthologous to another Arabidopsis cytosolic tyrosine aminotransferase, AtTAT2 (At5g36160) (Prabhu and Hudson 2010), was overexpressed approximately 2- to 3-fold in leaves silenced for NbPAT (de la Torre et al. 2014). The T-DNA-mediated suppression of ADT2 expression involves the concomitant suppression of the ADT and PDT enzymatic activities housed by this enzyme (Fig. 6). Our results suggest that the introduction of monofunctional PDT activity through the expression of the yeast PHA2 can rescue the adt2 mutant phenotype when expressed under control of the ADT2 promoter region and targeted to the plastids (Figs. 5, 6b). This result suggests that the phenylpyruvate pathway could be sufficient to supply Phe for seed development, at least when the arogenate pathway is blocked. The kinetic characterization of PHA2 (Supplementary Table S2) shows that this protein has roughly 20 times higher affinity towards prephenate and 700–800 times higher PDT activity than ADT2 (K_m = 680 µM, V_max = 1.6 pKat µg^–1; Cho et al. 2007). It provides evidence that under similar expression levels, PHA2 would boosts PDT if compared with ADT2, and even compete with PAT for their common substrate, prephenate.

Fig. 6

Model summarizing the complementation of ADT2 deficiency in Arabidopsis seed formation. This model highlights the essential role of the bifunctional ADT/PDT enzyme ADT2 for the viability of the embryo. (a) Seeds deficient in ADT2 are unable to synthesize phenylalanine or its derivatives, and therefore result in embryo arrest and seed abortion. When ADT activity was restored by the expression of ADT3, the flux through the arogenate-mediated pathway restored Phe biosynthesis and overall seed viability. (b) The PDT activity provided by the yeast PDT enzyme PHA2 in the mutant adt2 background restores Phe biosynthesis through a phenylpyruvate-mediated pathway and, concomitantly, the viability of the seed. (c) ADT2, ADT3 or PHA2 targeted to the cytosol are not able to restore seed viability, supporting that phenylpyruvate provision for Phe biosynthesis has a plastidial origin. The absence of complementation when ADT2 or ADT3 are overexpressed in the cytosol also suggests that there is no source of cytosolic arogenate. Dotted line: phenylpyruvate transport mechanism from plastids to the cytosol; the hypothetical provision of cytosolic chorismate and the role of CM2 remain unclear. Ppa, prephenate; Agn, arogenate; Ppy, phenylpyruvate; Cha, chorismate; ADT, arogenate dehydratase; PDT, prephenate dehydratase; PAT, prephenate aminotransferase; TAT, tyrosine aminotransferase; CM2, chorismate mutase 2.

The functionality of the phenylpyruvate pathway also requires the co-expression of an aminotransferase able to convert phenylpyruvate into Phe (Fischer and Jensen 1987a) (Fig. 1). Interestingly, from the different loci predicted to encode putative aromatic amino acid aminotransferases in Arabidopsis, the use of phenylpyruvate as substrate to produce Phe has only been confirmed experimentally in AtTAT1 and AtTAT2, both with a cytosolic location (Lopukhina et al. 2001, Prabhu and Hudson 2010, Riewe et al. 2012, Wang et al. 2016). Nevertheless, the conversion of phenylpyruvate into Phe has also been suggested to occur within the plastids, by an as yet unknown protein (Wang et al. 2016). In this regard, the data available from the Arabidopsis embryo (Winter et al. 2007, Slane et al. 2014, Palovaara et al. 2017) reveal that AtTAT1 and AtTAT2 transcripts are present during seed formation, albeit the AtTAT2 transcript is the most abundant. This observation supports the compartmentalization model proposed by Yoo et al. (2013), and suggests the putative involvement of AtTAT2 in this pathway (Fig. 6). Previous reports have claimed the participation of cytosolic-located ADT3 as PDT in a putative cytosolic Phe pathway under certain circumstances (Warpeha et al. 2006, Para et al. 2016). Nonetheless, our results suggested that neither of the cytosol-targeted enzymes, including cytosolic ADT3, were able to rescue the seed abortion phenotype, even if the CaMV P35S promoter is used to overexpress these activities. This might indicate that, even in the presence of cytosolic CM2, the provision of prephenate in the cytosol is not enough to restore Phe levels, or that this prephenate is unavailable due to metabolic channeling (for instance, through the formation of a metabolome) towards the synthesis of metabolites other than Phe. In summary, cytosol-targeting experiments in the adt2 background support that the provision of cytosolic phenylpyruvate for TAT enzymes rely on export from plastids to the cytosol, rather than on an intrinsic source in this last compartment.

Since ADT2 exhibits PDT activity in vitro and in vivo (Cho et al. 2007, Bross et al. 2011), we were interested in determining if the essential role of ADT2 in the embryo transition from the globular stage into the heart stage could be specifically related to its PDT activity. Nevertheless, our results demonstrate that the expression of the monofunctional ADT enzyme ADT3 in P_ADT2-ADT3 transgenic plants also results in a restored phenotype (Figs. 5,6). Taken together, our results suggest that the individual reintroduction of either of the two enzymatic activities of ADT2, ADT or PDT, recovers the adt2 mutant phenotype. Thus, it makes it possible that the biosynthesis of Phe during the seed formation in Arabidopsis would take place from both arogenate and phenylpyruvate. In turn, these observations may indicate that the essential role of ADT2 in Arabidopsis reproduction would be more linked to a specific gene expression pattern during the seed developmental program, rather than to specific enzymatic properties of ADT2. In this sense, ADT1, which also encodes a bifunctional ADT/PDT enzyme, is expressed in developing embryos of Arabidopsis (Supplementary Figs. S4, S7). However, the characteristic adt2 phenotype indicates that ADT1 does not play a redundant role with that of ADT2, reinforcing the hypothesis that the essentiality of ADT2 should be associated with its expression in a specific tissue or cell type. ADT2 has been proposed to have a ‘moonlighting function’ as a participant in the chloroplast division machinery (Bross et al. 2017) based on its specific subcellular location in a ring around the equatorial plane of chloroplasts. In this work, the authors reported that plants homozygous for the adt2-1D point mutation, which disrupts the ADT2 feedback inhibition mechanism in response to Phe (Huang et al. 2010), present misshapen and heterogeneous chloroplasts, suggesting that ADT2 could be critical for the plant for this reason. Nevertheless, our adt2 complementation studies with ADT3 and PHA2 do not support such statements as an explanation for the critical role of ADT2. If the moonlighting role suggested for ADT2, as part of the chloroplast division machinery, could determine the essentiality of this protein, neither ADT3 nor PHA2, which do not show the specific location described for ADT2 (Bross et al. 2017 for ADT2 and ADT3 and Supplementary Fig. S9 for PHA2), would be able to complement the mutation. Moreover, confocal imaging of our adt2 knock-out lines complemented with ADT2, ADT3 or PHA2 did not show any significant evidence of aberrant chloroplast morphology, as would be expected according to Bross and collaborators (Supplementary Fig. S10). Abnormal chloroplast morphology in the adt2-1D mutants could most probably be a consequence of an overaccumulation of Phe (up to 160-fold higher than in wild-type plants; Huang et al. 2010), as in fact not discarded by Bross et al. (2017).

The single-copy PAT-coding gene also seems to be essential for embryo formation (Pagnussat et al. 2005), initially suggesting that, although the phenylpyruvate pathway could be operative, the biosynthesis of Phe occurs mainly through the arogenate pathway. Nevertheless, pat and adt2 embryonic arrest most probably have a different developmental source, as PAT deficiency has been linked with a defect in female gametophyte formation (Pagnussat et al. 2005), whereas the female gametophytes from the adt2 heterozygous plants are morphologically normal (Supplementary Fig. S4). Furthermore, it is important to note that PAT is probably an essential enzymatic activity also for Tyr biosynthesis, since a prephenate dehydrogenase-coding gene (PDH; Fig. 1) has not been described to date in Arabidopsis (Schenck et al. 2015). Yet, pat lethality could be simultaneously linked to a deficiency in Tyr production, which lacks an alternative biosynthetic pathway in most plants. In contrast, in the absence of PAT, Phe could be produced from phenylpyruvate by the PDT activity carried by ADT2.

Materials and Methods

cDNA cloning

Arabidopsis thaliana ADT2 and ADT3 ORFs were cloned from around 50 ng of cDNA obtained from adult rosette leaves using the PCR primers ADT2Fwd/ADT2Rvs and ADT3Fwd/ADT3Rvs, respectively. The 5’-upstream region of ADT2 (P_ADT2) was cloned from genomic DNA of Arabidopsis extracted from mature rosette leaves. The primers PADT2Fwd and PADT2Rvs were used for this PCR. PHA2 was amplified from genomic DNA of S. cerevisiae strain BY4741 using the PCR primers ctpPHA2Fwd and PHA2Rvs. Based on Cho et al. (2007) and Bross et al. (2011), a 243 bp fragment encoding the putative N-terminal plastid transit peptide from ADT2 was amplified using the primers ADT2Fwd and ctpPHA2Rvs. Both PCR products were purified and fused to each other by PCR using the primers ADT2Fwd and PHA2Rvs. PCR products were purified, cloned into pJET1.2/blunt (Thermo Scientific) and confirmed by sequencing. All primer sequences are given in Supplementary Table S3.

Plasmid DNA constructs

Arabidopsis thaliana ADT2 was amplified from the pJET1.2::ADT2 construct using the primers PADT2ADT2Fwd and ADT2AttB2. P_ADT2 was amplified from the pJET1.2::P_ADT2 construct using the primers PADT2AttB1 and PADT2ADT2Rvs. Both PCR products were purified and fused using the primers PADT2AttB1 and ADT2AttB2. The resulting PCR product was re-amplified using the primers AttB1 and AttB2, purified from the gel and cloned into Gateway^® pDONR207^™ (Thermo Fisher Scientific) donor vector using BP Clonase^® II enzyme mix. Subsequently, it was recombined into the pGWB610 destination vector using the Gateway^® LR Clonase^® II mix.

To generate the PHA2 expression construct, the PHA2 ORF was amplified from the pJET1.2:PHA2 construct using the oligonucleotides pADT2ADT2Fwd and PHA2Rvs, and purified from the gel. In a second PCR, the primers PADT2AttB1 and PHA2AttB2 were used to fuse the PHA2 PCR product to P_ADT2, which was previously amplified from the pJET1.2::P_ADT2 construct using the primers PADT2Fwd and PADT2ADT2Rvs. The resulting PCR product was purified and re-amplified using the primers AttB1 and AttB2. This product was finally purified and cloned first into pDONR207^™, and then into the pGWB5 and pGWB610 destination vectors, as previously described.

Arabidopsis thaliana ADT3 was amplified from the pJET1.2::ADT3 construct using the primers PADT2ADT3Fwd and ADT3AttB2. P_ADT2 was amplified from the pJET1.2::P_ADT2 construct using the primers PADT2AttB1 and PADT2ADT3Rvs. Both PCR products were purified and fused using the primers PADT2AttB1 and ADT3AttB2. The resulting PCR product was re-amplified using the primers AttB1 and AttB2, purified from the gel and cloned into the Gateway^® pDONR207^™ (Thermo Fisher Scientific) donor vector using BP Clonase^® II enzyme mix. Subsequently, it was recombined into the pGWB610 destination vector, as described for pGWB610::P_ADT2-ADT2.

PHA2 without the plastid transit peptide (cytPHA2) was amplified using the primers PADT2PHA2Fwd and PHA2Rvs. The primers PADT2AttB1 and PHA2AttB2 were used to fuse the cytPHA2 PCR product to P_ADT2, previously amplified using the primers PADT2Fwd and PADT2PHA2Rvs. For the cytosolic versions of ADT2 (cytADT2) and ADT3 (cytADT3) under control of P_ADT2, the ORFs were amplified with the primers PADT2cytADT2Fwd/ADT2Rvs or PADT2cytADT3Fwd/ADT3Rvs, respectively. Forward primers were designed downstream of the putative transit peptides, whose approximate length was determined based on previous reports by Cho et al. (2007) and Bross et al. (2011). Fragments were fused to P_ADT2, previously amplified with the primer PADT2Fwd as forward and the corresponding reverse PADT2cytADT2Rvs or PADT2cytADT3Rvs. pGWB610 was used as destination vector. CaMV 35S-driven versions of cytPHA2, cytADT2 and cytADT3 were produced by amplifying the ORFs with the primer pairs cytPHA2AttB1/PHA2AttB2, cytADT2AttB1/ADT2AttB2 and cytADT3AttB1/ADT3AttB2. In a second round of PCR, Gateway sites were completed using the primers AttB1/AttB2. pDONR207^™ was used as entry vector. As destination vectors, pGWB11 was used for adt2 complementation in Arabidopsis, and pGWB5 to corroborate the cytosolic location of cytPHA2–GFP, cytADT2–GFP and cytADT3–GFP in N. benthamiana leaves (Supplementary Fig. S9).

P_ADT2 was fused to the GUS reporter gene (P_ADT2-GUS) amplifying P_ADT2 from the pJET1.2::P_ADT2 construct with the primers PADT2AttB1 and PADT2AttB2. This PCR product was re-amplified using the primers AttB1 and AttB2, purified from the gel and cloned into pDONR207^™. Later, it was recombined into the pGWB3 destination vector.

PCRs were carried out using the iProof^™ High-Fidelity Master Mix (Bio-Rad Laboratories). Escherichia coli DH5α was used as host strain. All primer sequences are given in Supplementary Table S3.

adt2 mutant complementation

Electrocompetent cells of the Agrobacterium tumefaciens C58C1 strain were transformed with the pGWB expression constructs and selected for the corresponding antibiotic resistance. Heterozygous adt2-1 plants (T₀) were transformed using the floral dip method (Clough and Bent 1998). Control plants were transformed with the corresponding empty vectors.

T₁ seeds were selected for dl-phosphinothricin (Duchefa Biochemie) resistance in 0.5 × Murashige and Skoog (MS) medium. Resistant plants were transferred into soil. The transgenic construct and the adt2-1 T-DNA insertion were confirmed by PCR using the primers AttB1/AttB2 and 044Fwd/LBTDNA, respectively. Three transgenic lines were selected for each construct and control empty vectors, and their T₂ seeds were selected for dl-phosphinothricin resistance in 0.5 × MS plates. Thirty plants for each transgenic line were transferred into soil, and analyzed by PCR for homozygous T-DNA insertion. T₃ plants descended from homozygous adt2-1 T₂ plants were selected for dl-phosphinothricin resistance; the homozygous adt2-1 genotype was subsequently confirmed in all the plants of the T₃ progeny. Total RNA extraction from T₃ plants and RT–qPCR analysis were performed as described by de la Torre et al. (2014). Primer pair q-ADT2Fwd/q-ADT2Rvs, designed on the non-coding 3’-untranslated region of ADT2, absent in the P_ADT2-ADT2 construct, were used selectively to detect endogenous ADT2 expression. The expression level of elongation factor 1α (primer pair EF1aFwd/EF1aRvs) was used for normalization (Supplementary Table S3).

Embryo visualization

Arabidopsis adt2-1 and adt2-2 siliques in different developmental stages were collected and cleared in Herr’s solution (Liu et al. 2008) prior to imaging with a Leica TL3000 Ergo stereo-microscope and a Nikon Eclipse Ti-E Inverted Microscope (Nikon Instruments). All pictures were processed with the Fiji distribution of ImageJ (Schindelin et al. 2012).

GUS staining

Siliques in different developmental stages from three independent pGWB3::P_ADT2-GUS transgenic lines were taken and infiltrated with the GUS staining solution using a vacuum pump. Infiltrated samples were incubated overnight at 37°C in the dark. After GUS incubation, samples were incubated in ethanol:acetic acid 1:1 for 4 h and destained in Hoyer’s medium for 2 d before imaging (Stangeland and Salehian 2002). GUS staining solution was prepared as described by Kirchsteiger et al. (2012). Transgenic Arabidopsis lines carrying the empty vector pGWB3 were employed as controls.

Accession numbers

Sequence data can be found in the GenBank/EMBL or Arabidopsis Genome Initiative (AGI) databases: ADT2 (At3g07630), ADT3 (At2g27820) and PHA2 (CAA86380.1).

Funding

This work was supported by the Spanish Ministerio de Economía y Competitividad [grant No. BIO2015-69285-R]; Junta de Andalucía [grant No/ BIO-474]; and the Ministerio de Educación, Spain [a Formación Profesorado Universitario fellowship to J.E.A.].

Acknowledgments

We would like to thanks the students Ana Álvarez, Lorena Aguilera and Alberto Urbano for their helping on processing the very large number of Arabidopsis genomic DNA extractions that this study has required. We are also grateful to Dr. John Pearson (Bionand, Málaga, Spain) for his technical assistance with confocal microscopy, and to Dr. Belén Pascual (University of Málaga, Spain) for her advice on GUS staining. pGWB610 vector was kindly provided by Tsuyoshi Nakagawa (Shimane University, Matsue, Japan). F.T. and J.E.A. conceived the project and research plans; F.T., J.E.A. and F.M.C. designed the experiments and analyzed the data with collaboration from C.A.; J.E.A. performed most of the experiments with collaboration from F.T.; J.E.A. wrote the article with contributions from all authors. F.M.C. and C.A. were responsible for funding acquisition and project administration.

Disclosures

The authors have no conflicts of interest to declare.