Deregulation of phenylalanine biosynthesis evolved with the emergence of vascular plants

Abstract

Phenylalanine (Phe) is the precursor of essential secondary products in plants. Here we show that a key, rate-limiting step in Phe biosynthesis, which is catalyzed by arogenate dehydratase, experienced feedback de-regulation during evolution. Enzymes from microorganisms and type-I ADTs from plants are strongly feedback-inhibited by Phe, while type-II isoforms remain active at high levels of Phe. We have found that type-II ADTs are widespread across seed plants and their overproduction resulted in a dramatic accumulation of Phe in planta, reaching levels up to 40 times higher than those observed following the expression of type-I enzymes. Punctual changes in the allosteric binding site of Phe and adjacent region are responsible for the observed relaxed regulation. The phylogeny of plant ADTs evidences that the emergence of type-II isoforms with relaxed regulation occurred at some point in the transition between nonvascular plants and tracheophytes, enabling the massive production of Phe-derived compounds, primarily lignin, a hallmark of vascular plants.

The appearance of arogenate dehydratase isoenzymes with reduced feedback inhibition during the early evolution of plants enabled the massive biosynthesis of phenylalanine-derived compounds.

Introduction

Aromatic amino acids (AAAs) phenylalanine (Phe), tyrosine (Tyr), and tryptophan (Trp) are of fundamental importance for all forms of life. However, only bacteria, fungi, and plants have the necessary biochemical pathways for the biosynthesis of these amino acids (Maeda and Dudareva, 2012). In accordance with their importance, evolution has triggered regulatory mechanisms of AAA biosynthesis at the transcriptional and posttranscriptional levels, enabling a fine control of the metabolic flux through these pathways. This is particularly important in plants, where AAAs serve as precursors for the biosynthesis of a wide range of natural compounds including phenylpropanoids, alkaloids, indole auxins, and betalains (Schenck and Maeda, 2018). Some of these downstream products, such as lignin, account for a large extent of the total plant biomass (Tohge et al., 2013). An adequate provision of precursors will be necessary to maintain the production of such specialized metabolites.

In plants, the biosynthesis of Phe occurs through two alternative routes (Figure 1A). In the arogenate (Agn) pathway, prephenate (Ppa) is transaminated by Ppa-aminotransferase to generate Agn, which is decarboxylated and dehydrated by Agn dehydratase (ADT) to give Phe (Bonner and Jensen, 1987). Alternatively, in the phenylpyruvate pathway, Ppa is converted first into phenylpyruvate by Ppa dehydratase (PDT), which is the substrate of phenylpyruvate aminotransferase, producing Phe (Yoo et al., 2013). This last pathway has been confirmed to be cytosolic in plants (Qian et al., 2019). The Agn pathway has been reported to be responsible for the main provision of Phe (Maeda et al., 2010, 2011), although various investigations have reported the contribution of the phenylpyruvate pathway (Yoo et al., 2013; Oliva et al., 2017; El-Azaz et al., 2018; Qian et al., 2019). Furthermore, Agn is a precursor for the biosynthesis of Tyr through the action Agn dehydrogenase (ADH/TyrA_a) (Bonner and Jensen, 1987; Fischer and Jensen, 1987b, reviewed by Schenck and Maeda, 2018).

Figure 1.

Effector-mediated regulation of the biosynthesis of aromatic amino acids in plants. A, 3-Deoxy-D-arabino-heptulosonate-7-phosphate synthase, AS, CM, ADH, and ADT/PDT are targets of different feedback-regulatory loops. E4P, erythrose 4-phosphate; PEP, phosphoenolpyruvate. B, Structure of plant ADTs include an N-terminal putative chloroplast transit peptide, a catalytic domain and a C-terminal ACT regulatory domain, in which the PAC domain comprised (C) Allosteric regulation of ADT by Phe is mediated by a conformational change in the enzyme, postulated as a homotetramer in its quaternary structure. Phe binds to the ACT domain in a pocket formed in the conjunction of the ACT domains from two enzymes monomers. The binding of Phe promotes the reversible transition of the whole enzyme to a T-state, in which the accessibility of the active site is reduced (adapted from Tan et al., 2008).

In addition to transcriptional regulation, some key enzymes of AAA biosynthesis, including ADT, are subjected to effector-mediated regulation mechanisms that determine flux allocation into different branches of the shikimate pathway (Maeda and Dudareva, 2012; Figure 1A). ADTs from plants belong to a family of enzymes that are composed of an N-terminal cyclohexadienyl dehydratase catalytic domain fused to Aspartate kinase, Chorismate mutase, and TyrA (ACT) regulatory domain (Figure 1B) also present in other enzymes regulated in response to amino acid levels. Cyclohexadienyl dehydratases have the potential of using Ppa and l-Agn as alternative substrates (Xia et al., 1991; El-Azaz et al., 2016; Clifton et al., 2018), being the superior efficiency in the use of one or other substrate the cause for the enzyme name. These enzymes are typically tetramers (dimers of dimers): dimerization is mediated by the interaction between catalytic and regulatory domains of the monomers, whereas the tetramer is formed only by ACT–ACT contacts (Tan et al., 2008). The ACT domain mediates in the feedback inhibition of the enzyme by Phe, by inducing a conformational change that makes the active site inaccessible to the substrate (Figure 1C; Jung et al., 1986; Pohnert et al., 1999; Tan et al., 2008). Previous investigations have demonstrated that the mutation of the residues involved in the allosteric binding of Phe results in feedback insensitive ADTs, thereby promoting the accumulation of very high levels of Phe in rice (Oryza sativa) and Arabidopsis (Yamada et al., 2008; Huang et al., 2010). The severe effect of ADT-deregulation highlights the important role of effector-mediated regulation in maintaining Phe homeostasis.

Cho et al. (2007) performed a classification of ADTs considering several bacterial species as well as two plant species, rice and Arabidopsis thaliana. In a previous work (El-Azaz et al., 2016), we deepened the phylogenetic classification of ADTs and proposed a distribution of ADT enzymes into four Subgroups (I–IV). Subgroups I and II include all ADTs from flowering plants, whereas Subgroups III and IV were only found to be present in gymnosperms and green algae, respectively. Interestingly, our sequence analysis showed that the protein domain involved in the allosteric regulation by Phe has remarkable differences between plant lineages. Hence, within flowering plants, isoforms from Subgroups I and II singularly differ in the presence (Subgroup I) or absence (Subgroup II) of a 21 amino acid region named PAC (from PDT Activity Conferring), that overlaps the allosteric Phe binding site and several residues from the tetramerization interface in the ACT domain (Figure 1B). This region also corresponds with the ability to complement PDT deficiency in the yeast auxotrophic mutant pha2 (El-Azaz et al., 2016). Enzymes within subgroup III seem to be found only in conifers and some groups of flowering plants and, to date, remains unclear how long these enzymes are bona fide ADT or PDT (El-Azaz et al., 2016). Subgroup IV contains enzymes from green algae that also have a PAC domain.

Based on these previous evidence, we hypothesized that the different groups of ADTs may have undergone substantial changes in the allosteric regulation by Phe. Our findings support that ADTs within Subgroup I and IV (hereafter type-I) strongly differ from Subgroup II (hereafter type-II) in their sensitivity to feedback-inhibition by Phe. Type-I enzymes, which are common to all land plants and algae ancestors, exhibit a tight inhibition by Phe. Conversely, type-II enzymes show a considerably lesser degree of inhibition by Phe and are only found in euphyllophytes (ferns and seed plants).

Phylogenetic studies of a large number of sequences from lycophytes and ferns support that type-II ADTs diverged from a preexistent gene duplicate of a type-I isoform in the ancestors of modern vascular plants, probably as an adaptation to the massive demand of lignin and other Phe-derived compounds. Taken together, these findings provide insights into the biochemical regulation and evolution of Phe biosynthesis in land plants, with possibilities for future biotechnological applications.

Results

Two ADTs from maritime pine, Pinus pinaster, exhibit radical differences in their sensitivity to Phe inhibition

As mentioned above, ADT activity in plants has long been known to be subjected to feedback-inhibition by the endproduct of the reaction, Phe (reviewed by Maeda and Dudareva, 2012). Previous works performed using crude extracts from plant tissues, or recombinant proteins provided limited information on how this regulatory mechanism affects ADT isoforms (Jung et al., 1986; Chen et al., 2016). Using representative members of the main ADT subgroups in maritime pine, PpADT-G, from subgroup I and PpADT-C from subgroup II, we proceeded to test putative differences in the Phe feedback inhibition dynamic of ADT activity.

Consistently with previous reports (Jung et al., 1986), we observed that PpADT-G is inhibited by micromolar levels of Phe. The concentration of Phe that causes a 50% of inhibition of the enzyme (IC₅₀) was calculated at 27.6 µM of Phe (Figure 2A andTable 1). Conversely, no substantial decrease in ADT activity was observed for PpADT-C activity up to 100 µM of Phe (Figure 2A), a condition at which PpADT-G was found to be mostly inactive. Additional kinetical parameters for PpADT-G and PpADT-C as ADT are summarized in Supplemental Figure S1 and Supplemental Table S1.

Figure 2.

Alternative ADT isoforms have different sensitivities to Phe as negative effector. A, Inhibition of ADT activity by Phe in the wild-type enzymes PpADT-G and PpADT-C. B, Inhibition of PDT activity in PpADT-G and PpADT-C. C, Scheme illustrating the swapping of the PAC domain between PpADT-G and PpADT-C. D, Inhibition of PDT activity in the chimeric enzymes PpADT-Gmut¹ and PpADT-Cmut¹. E, Inhibition of ADT activity in PpADT-Gmut¹ and PpADT-Cmut¹. Errors bars represent sd from three independent replicates.

To further characterize this differential response to Phe, we decided to take advantage of the PDT activity exhibited by both enzymes, bypassing the technical limitation that the addition of large amounts of Phe represents for accurate determination of ADT activity. Assayed as PDT at 1 mM of Ppa, PpADT-G reached IC₅₀ at 47.7 µM of Phe (Figure 2B andTable 1). Kinetics analysis indicated that the inhibition mechanism of PpADT-G by Phe is apparently uncompetitive, which implies that Phe only can bind to the enzyme when the enzyme–substrate complex is already established (Supplemental Figure S2). The affinity constant for Phe (K_i) was estimated to be 28.1 µM. In contrast, PpADT-C was found to be inhibited only over 100 µM of Phe, with an estimated IC₅₀ value of 320 µM (Table 1). In the absence of Phe, PpADT-C exhibited a Michaelian response to substrate concentration as illustrated in Supplemental Figure S2C. In the presence of Phe above 200 µM, this enzyme apparently switches to a sigmoidal kinetic, typical of allosteric regulation (Supplemental Figure S2C; Palmer, 1995). However, this result might have been impacted by the difficulty to accurately determine the activity of this enzyme at low substrate concentrations in the presence of inhibitory concentrations of Phe.

Table 1.

Kinetic parameters of Phe-mediated inhibition in PpADT-G, PpADT-C, PpADT-Gmut¹, and PpADT-Cmut¹

Enzyme	ADT	PDT
	IC50 (μM of Phe)	IC50 (μM of Phe)	K i (μM of Phe)	Proposed type of inhibition
PpADT-G	27.6 ± 1.5	47.7 ± 3.0	28.1 ± 1.0	Uncompetitive
PpADT-C	No substantial inhibition up to 100 μM of Phe	320.2 ± 10.7	ND	Cooperative
PpADT-Gmut1	No substantial inhibition up to 100 μM of Phe	389.9 ± 15.4	ND	Cooperative
PpADT-Cmut1	52.8 ± 2.9	36.1 ± 3.4	170.5 ± 16.3	Uncompetitive

Parameters are expressed as an average of three independent replicates ± sd. IC₅₀ values were determined at 0.1 mM of Agn (for ADT activity) or 1 mM of Ppa (for PDT activity).

Differences in the Phe binding region determine the relaxed regulation of PpADT-C

PpADT-G and PpADT-C differ in the sequence of the regulatory Phe binding site and oligomerization interface within the ACT domain. To address whether such changes in the sequence explain the different levels of inhibition by Phe, a domain swapping of this sequence motif (Figure 2C) was performed between PpADT-G and PpADT-C, resulting in two chimeric enzymes: PpADT-Cmut¹, which contains the PAC domain from PpADT-G, and PpADT-Gmut¹, its reciprocal counterpart. The response to Phe in the chimeric enzymes was changed compared to their wild-type versions (Figure 2, D and E).

PpADT-Cmut¹ exhibited uncompetitive inhibition (Supplemental Figure S2), with estimated IC₅₀ values of ∼52.8 µM and ∼36.1 µM (assayed, respectively, as ADT or PDT; Figure 2, D and E; Table 1), resembling PpADT-G. The K_i parameter for Phe was estimated to be 170.5 µM, which was markedly higher than that of PpADT-G.

The reciprocal mutant enzyme PpADT-Gmut¹ exhibited relaxed regulation in response to Phe (Figure 2, D and E; Table 1). No substantial inhibition of ADT activity was observed up to 100 µM of Phe, and the IC₅₀ value was determined to be almost 400 µM. Moreover, it was also corroborated that the inhibition mechanism changed from uncompetitive, for PpADT-G, to allosteric, for PpADT-Gmut¹ (Supplemental Figure S2).

Deregulated ADTs are widespread in seed plants

Phylogenetic analysis of the ACT domain of ADTs from seed plants indicates that this domain is distributed into two groups in all vascular plants analyzed (Figure 3A). ACT domains that clustered together with the ACT domain from PpADT-G, which contains the PAC sequence, were named type-I. Conversely, those containing the ACT domain from PpADT-C were named type-II, and included enzymes lacking the PAC sequence. Two additional clusters of putative ADTs, integrated by sequences from conifers and monocots that do not correspond to the features of type-I and -II isoforms, were also identified. These additional clusters include sequences previously described as subgroup III (El-Azaz et al., 2016). We hypothesized that this distribution could correspond to the existence of regulated (type-I) and deregulated (type-II) ADT isoforms among other spermatophytes beyond conifers.

Figure 3.

ADTs from seed plants are distributed into two common groups, type-I and -II, that differ in the level of regulation by Phe. A, Neighbor-joining phylogenetic analysis of the ACT domain of ADTs from seed plants. Type-I, corresponding to putatively tight regulated ADTs, is marked in green. Type-II, corresponding to ADTs with relaxed regulation, is marked in yellow. Tree was set unrooted. Confidence probability is expressed in percentage and was estimated using the bootstrap test (1,000 replicates). The scale bar corresponds to 0.2 estimated amino acid substitutions per site. Deletion of ambiguous position was set up at a conservation rate of 90%. Species abbreviations: Ambtr, A. trichopoda; Aquca, Aquilegia caerulea; At, A. thaliana; Bradi, Brachypodium distachyon; Cucsa, C. sativus; Eucgr, Eucalyptus grandis; Glyma, Glycine max; Niben, N.benthamiana; Orysa, O. sativa; Pethy, Petunia hybrida; Picab, Picea abies; Potri, P. trichocarpa; Pp, P. pinaster; Zeama, Z. mays. B, Effect of 100 μM Phe over ADT activity in recombinant type-I ADTs from different plants (names in blue in (A)). ADT activity is expressed in relative units (control without Phe = 100% activity). C, Effect of Phe over type-II recombinant ADTs (names in blue in (A)). ADT activity was determined in triplicate at 100 µM of substrate (Agn). For (B) and (C) values are means of three replicas (n = 3) and errors bars represent sd.

To contrast this hypothesis, we identified and cloned ORFs encoding four type-I ADTs from distinct species of flowering plants: A. thaliana (AtADT1), Populus trichocarpa (Potri11G4700), Nicotiana benthamiana (Niben8991), and Cucumis sativus (Cucsa52640). Such proteins were recombinantly produced, and ADT activity was determined at an initial concentration of Phe (100 μM) and compared to the control (absence of Phe; Figure 3B). The four type-I enzymes exhibited a strong decrease in ADT activity (>90%) at 100 µM of Phe, similar to the type-I enzyme PpADT-G. In parallel, we performed the same experiment with four type-II enzymes: AtADT4 (from A. thaliana), Potri4G188100 (from P. trichocarpa), Orysa4G33390 (from O. sativa), and Zeama2G125923 (from Zea mays). The ADT activity exhibited by these type-II enzymes was not substantially affected at the concentration of Phe used in the assay (Figure 3C), confirming our previous observations in PpADT-C and demonstrating that deregulated ADTs are widespread in flowering plants.

Overexpression of type-II ADTs has a major impact on Phe levels in planta

To test the physiological significance of allosteric deregulation of type-II ADTs, we determined how the overexpression of type-II enzymes would impact Phe accumulation in planta compared to type-I enzymes. Leaves from N. benthamiana plants overexpressing type-I enzymes PpADT-G, AtADT1, and AtADT-2, respectively, accumulated Phe to an average level of 4, 32, and 6 times compared to control (GFP). In contrast, the Phe levels in leaves overexpressing type-II enzymes PpADT-C and PpADT-A were found to be approximately 145 and 160 times higher compared to controls, and up to 40 times higher compared to the type-I enzymes (Figure 4). The levels of the enzymes were determined by western blotting in the same samples (Supplemental Figure S3), indicating that the large differences observed in Phe content cannot be attributed to a higher expression level of type-II enzymes.

Figure 4.

Phe accumulation in plants overexpressing regulated or deregulated ADTs. Errors bars represent sd from eight independent replicates. Different letters indicate significant differences in the Student’s t test (p-value = 0.01). Species abbreviations: Pp, P. pinaster; At, A. thaliana; GFP, green fluorescent protein.

To further support the essential role of the PAC domain in the allosteric regulation of ADTs, the chimeric enzymes PpADT-Gmut¹ and PpADT-Cmut¹ were included in this experiment. Precedent in vitro characterization indicates that the relaxed regulation by Phe can be exchanged by swapping the PAC domain. Consistently, leaves overexpressing PpADT-Cmut¹ accumulated Phe to a level comparable to leaves overexpressing wild-type PpADT-G (Figure 4). The estimation of the protein levels of PpADT-Cmut¹ by western blot analysis indicated that, in this regard, no major differences could be found when compared to the wild-type version (Supplemental Figure S3). Overall, differences in Phe accumulation are likely a consequence of different sensitivity of the enzymes to Phe as a negative effector.

Various residues within or adjacent to the Phe binding pocket in the ACT domain determine a lower sensitivity to Phe as a feedback-inhibitor

Phylogenetic analysis differentiates two groups of ADTs in seed plants, with an alternative sequence in the region that binds Phe to the enzyme’s regulatory domain (the ACT domain). In type-I isoforms, Phe biding site typically contains the signature residues Thr/Ser³⁰³, Leu³⁰⁴, Pro³⁰⁸, Gly³⁰⁹, Ala³¹⁴, Ala³¹⁶, Val³¹⁷, Leu³²⁰, and Asn³²⁴, with few exceptions (positions as in PpADT-G; Supplemental Figure S4). In contrast, Ala³⁰³, His/Gln³⁰⁴, Thr³⁰⁸, Ser³⁰⁹, Val³¹⁴, Ser³¹⁶, Ala³¹⁷, Phe³²⁰, and Ser³²⁴ are the most common residues in type-II enzymes (Supplemental Figure S4). Figure 5A depicts a logo sequence of this region that summarizes such differences between type-I and -II ADTs.

Figure 5.

Decreased sensitivity to feedback-inhibition by Phe is determined by various residues in the Phe binding region. A, Type-I (green) and type-II (yellow) enzymes differ in the primary sequence of the Phe binding pocket region within the ACT domain. Purple color indicates residues that are preserved between both groups; black color corresponds to barely conserved residues. B, Site-directed mutagenesis affecting PpADT-G, replacing residues that are highly conserved in type-I isoforms (green) by the corresponding residues from type-II isoforms (yellow). C, Determination of the IC₅₀ parameter in the mutant versions of PpADT-G with decreased sensitivity to Phe as a negative effector, compared to wild-type. Measurements were done in triplicate (n = 3) at a concentration of Ppa of 1 mM. D, Effect of 100 μM Phe over ADT activity in recombinant PpADT-G mutant proteins. Error bars represent sd.

Next, we performed a phylogeny-guided site-directed mutagenesis study using PpADT-G. Differences in the consensus sequence between type-I and type-II isoforms were used to generate 11 mutant versions of PpADT-G (Figure 5B). The sensitivity to Phe inhibition, determined as the IC₅₀ value for the PDT reaction, was determined for each mutant version of the enzyme (Table 2). IC₅₀ was found to be substantially increased in the mutant proteins PpADT-Gmut⁶, Gmut⁶², Gmut⁸, Gmut¹⁰, Gmut¹⁰¹, and Gmut¹⁰², ranging from 184 µM for PpADT-Gmut¹⁰ to 445 µM in PpADT-Gmut⁸ (Figure 5C). On the other hand, mutations affecting the residues Ala³¹⁴, Ala³¹⁶, Val³¹⁷, and their combinations (mutant proteins PpADT-Gmut⁹, Gmut¹¹, Gmut¹², Gmut¹³, and Gmut¹⁶) increased the sensitivity toward Phe as inhibitor (Supplemental Figure 5). Based on the observed effect over PDT activity, we selected some of the mutant versions of PpADT-G to determine the inhibitory effect of Phe on ADT activity. PpADT-Gmut⁶, Gmut⁸, and Gmut¹⁰ were assayed as ADT in the presence of Phe at 100 µM (Figure 5D), showing that PpADT-Gmut⁸ is more resistant to inhibition than the wild-type. Interestingly, PpADT-Gmut⁶ and PpADT-Gmut¹⁰ exhibited similar or higher sensitivity to Phe-mediated inhibition of ADT activity compared to the wild-type, suggesting that PDT and ADT activities harbored by the same protein might not be regulated in the same exact way. Together, these results support the relevance of the residues between positions 303 and 324 on the regulation of the enzyme’s activity, and remark the especial significance of residues at positions 308 and 309 for both ADT and PDT activities of PpADT-G.

Table 2.

IC₅₀ values for Phe in the mutant enzymes derived from PpADT-G

	Substitution(s)	IC50 (μM)
PpADT-Gmut6	Thr303 → Ala303 Leu304 → Gln304	332.8 ± 8.0
PpADT-Gmut62	Thr303 → Ala303 Leu304 → His304	333.5 ± 3.1
PpADT-Gmut8	Pro308 → Thr308 Gly309 → Ser309	444.8 ± 42.1
PpADT-Gmut9	Ala314 → Val314 Ala316 → Ser316 Val317 → Ala317	17.5 ± 3.7
PpADT-Gmut11	Ala314 → Val314	14.6 ± 2.3
PpADT-Gmut12	Ala316 → Ser316	26.0 ± 4.8
PpADT-Gmut13	Val317 → Ala317	23.6 ± 1.4
PpADT-Gmut16	Ala316 → Ser316 Val317 → Ala317	17.2 ± 0.2
PpADT-Gmut10	Leu320 → Phe320 Asn324 → Ser324	184.0 ± 24.7
PpADT-Gmut102	Leu320 → Phe320	135.4 ± 0.7
PpADT-Gmut103	Asn324 → Ser324	234.3 ± 4.9

IC₅₀ is expressed as an average of three independent replicates ± sd at 1 mM of substrate (Ppa).

In silico modeling provides structural support for the deregulation of type-II ADTs

To elucidate how the differences in the primary structure of the ACT domain can determine the allosteric response to Phe, PpADT-G, and PpADT-C 3D-structures were modeled by homology (Figure 6). The N-terminal region of the ACT domain between residues 303–324 in PpADT-G, and 332–353 in PpADT-C, encompasses the last three residues of the first β-strand of ACT, the following α-helix and the loops at both ends of the helix (Figures 6A;Supplemental Figure S6). In the relaxed form, the residues involved in the binding of Phe form four grooves open to the solvent at each edge of the ACT-dimerization interface (two by dimer). In the tense form, the grooves between ACT domains change to close cavities, and the ACT-interfaces between the dimers and tetramers show a tighter fit (Figure 6, A and B). In addition, the catalytic domains in each dimer are closer than in the full active form, changing the active site conformation and reducing the accessibility of substrate (Figure 6A;Tan et al., 2008).

Figure 6.

3D-modeling of the allosteric Phe cavity. A, Right, cartoon of PpADT-G R and T structural models. Tetramers are colored by chain (A–D), ACT regulatory domains are framed by a rectangle. Allosteric Phe in T model is shown as sticks. A/B and C/D dimers form a tetramer through ACT domain interactions. Rotation of ACT domains, restructuring of the ACT β-sheet in contact with the catalytic domains and active sites closure are remarkable conformational changes between R and T models. Left, detail of the ACT domains of PpADT-G in the tense conformation. The secondary structure elements of the domain are identified (β1α1β2β3α2β4 fold). In gray, the PAC region in each monomer, encompassing the β1 C-terminal end, β1/α1 connecting loop, α1, and half of α1/β2 connecting loop. The four allosteric Phe are shown in stick representation, two by dimer interface. B, Superposition between ACT dimers from tense (colored by chain as in A) and relaxed (gray) structural models, highlighting conformational differences. PAC regions from chains A and B would have to approach to form the cavities. C, Zoom over part B showing one of the Phe binding sites. The residues that have been mutated in the present work resulting in the increase of Phe IC₅₀ are represented as sticks in the tense (green and cyan) and relaxed (gray) conformations. D, Transparent surface detail of one of the allosteric cavities from PpADT-G (left) and PpADT-C (right) tense structural models, revealing the bound Phe represented as sticks. Residues forming the cavity from chains A and B are shown as green and cyan sticks, respectively. PAC residues from both chains are named with red letters, whereas black letters name cavity-forming residues out of the PAC region. PpADT-G residues that have been mutated to the equivalent residues in PpADT-C are underlined in both enzymes. Thr³⁰³ and Leu³²⁰ from PpADT-G (Ala³³² and Phe³⁵⁹ in PpADT-C) are outside but close to the Phe binding cavity. Discontinuous black lines represent polar interactions between residues of the cavity and the allosteric Phe backbone. Asn³²⁴Ser mutation in PpADT-G (Ser³⁵³ in PpADT-C) avoids the interaction with the backbone amino group of Phe and the cavity fails to close properly. Discontinuous red line shows the polar interaction between Ser³³⁸ and Ala³⁴⁸ in PpADT-C, which is not possible between the equivalent positions in PpADT-G. Amino acid one-letter code was used for naming the residues. Red and blue colors were used for oxygen and nitrogen atoms, respectively, in the stick representation.

According to our predictions, Ser³⁵³ in PpADT-C instead of Asn in the PpADT-G equivalent position (Asn³²⁴) avoids the polar interaction with the Phe amino group and, due to its smaller size, keeps two of the four Phe binding cavities slightly open to the solvent, which would be compatible with a reduction in affinity for Phe as inhibitor (Figure 6, C and D). In PpADT-Gmut¹⁰, this substitution was introduced together with Leu³²⁰Phe (Phe³⁵⁹ in PpADT-C). The interactions between large residues, such as Phe³¹² and Leu³²⁰ (Phe³⁴¹ and Phe³⁵⁹ in PpADT-C), are responsible for tetramerization of the monomers (Figure 6D). The transition from the relaxed into the tense state involves a decrease in the distance between such residues. Considering that Phe is larger than Leu, steric hindrance between the eight Phe rings placed in the ACT tetramer may restrict the concerted movement to reach the tense conformation (see Morph simulation at the Supplemental material). Therefore, both effects, reduction in the affinity for Phe and less efficient transition to the tense form, are in good agreement with the observed Phe IC₅₀ increase in Gmut¹⁰ when assayed as PDT. On the other hand, our results indicate that PpADT-Gmut¹⁰ is still highly sensitive to feedback inhibition when assayed as ADT, evidencing that the Phe-induced transition between tense and relaxed form could be determined by the enzyme substrate available.

Mutations placed at the N-terminal side of the PAC region (PpADT-Gmut⁶, Gmut⁶², and Gmut⁸) produce a substantial impact on de-regulation of only PDT activity (Gmut6 and Gmut62) or both ADT and PDT activities (Gmut8), even though any deleterious effect on the cavity shape or the interactions with the Phe backbone could be inferred from the structural models. Superposition of relaxed and tense models for PpADT-G and PpADT-C shows that residues in this region (303–309 and 332–338 in PpADT-G and -C, respectively) form a loop connecting the β-strand and the α-helix of the PAC region with a turn, experiencing strong positional and conformational changes in the transition to the tense state (Figure 6C). This finding suggests that the flexibility in this loop is important to fulfill this function in the ACT domains. Hence, considering that Gly is the amino acid with the smallest side chain, Gly³⁰⁹Ser substitution in PpADT-Gmut⁸ will reduce the flexibility of this loop. Interestingly, a recent study combining in silico dynamics simulation and experimental results has shown that this region is part of a ligand gate whose movements allow the entrance of the allosteric Phe to the cavity in the ACT domain of human Phe hydroxylase (Ge et al., 2018). Moreover, it has been found that the same mutation at the homologous position of human Phe hydroxylase (G46S) completely prevents the binding of Phe to the regulatory domain (Leandro et al., 2017). In contrast, P308T (the second mutation in PpADT-Gmut⁸) releases the rigidity in the peptidic backbone, and together with Gly³⁰⁷, it likely contributes to smoothing the effect of G309S substitution. In contrast, Thr³⁰⁸ (Thr³³⁷ in PpADT-C) is predicted to form a hydrogen bond with Ala³¹⁹ (Ala³⁴⁸ in PpADT-C) from the opposite monomer that forms the cavity (Figure 6D), stabilizing the conformation in the tense form. Taken together, our model indicates that P308T and G309S would counterbalance their individual effects to produce a loop that is less flexible but still functional.

Last, PpADT-Gmut⁶ and Gmut⁶² mutations affect Thr³⁰³ and Leu³⁰⁴ at the C-terminal end of the ACT first β-strand (Figure 6C). Thr³⁰³ does not form part of the predicted Phe cavity, whereas Leu³⁰⁴ is involved in a small portion of it. In the transition between relaxed and tense conformations, ACT β1 moves together with the loop comprised between positions 303–309 to form the cavity (Figure 6C; Morph simulation at Supplemental material), changing its backbone interactions with the adjacent strands of the β-sheet and the side chains. Although the T303A mutation does not appear to have an effect on the structure or the allosteric conformational change, L304Q or L304H introduce a voluminous side chain with a polar group between hydrophobic residues (Figure 6D), producing steric hindrance in the tense conformation, but not in the relaxed conformation. The active relaxed conformation would be favored over the tense conformation, in concordance with the increase of IC₅₀ of PpADT-Gmut⁶ and PpADT-Gmut⁶² when assayed as PDTs. As stated above for PpADT-Gmut10, the differential effect of these mutations on PDT and ADT reactions suggest that Agn and Ppa could impact the transition and/or stability of relaxed and tense states in a way that remains unclear.

Deregulated type-II ADTs are a distinctive feature of vascular plants

With the aim of addressing the distribution of deregulated ADT activity through the plant kingdom, we examined the occurrence of the mutations T303A, L304H/Q, P308T, G309S, L320F, and N324S, in the PAC domain of ADTs from green algae to flowering plants. We identified 72 nonredundant ADT sequences from green algae, 62 from liverworts, 87 from bryophytes, 78 from lycophytes, 425 from pteridophytes, and 69 from spermatophytes. Partial sequences included at least the complete studied region. Figure 7 summarizes the relative distribution of residues in the Phe binding site and adjacent residues, showing that approximately half of the ADT isoforms from seed plants likely correspond to enzymes recalcitrant to feedback inhibition. Remarkably, the combination of residues that reduce sensitivity to Phe inhibition is completely absent or uncommon in the enzymes from algae, nonvascular plants (liverworts and bryophytes), and lycophytes (Selaginella moellendorfii). In contrast, ADT isoforms from pteridophytes already exhibit the entire set of residues that characterize the Phe binding region of type-II isoforms, except for N324S, which is rare. Additionally, these residues were found to co-occur within the same isoforms from ferns, similar to seed plants (Supplemental Figure S4). The frequency of co-occurrence between these residues for both type-I and type-II enzymes has been calculated, by using this set of multiple sequences from different groups of plants (Supplemental Dataset 1). This analysis, support our previous observations regarding the co-occurrence of many of them (Figure 7). In Supplemental Figure S7, we illustrate the co-ocurrence of a set of amino acids characteristic of type-I enzymes and involved in the feedback inhibition (Ser/Thr303, Leu304, Pro308, Gly309, Leu 320, and Asn324).

Figure 7.

Conservation frequencies in percentage of critical residues determining high sensitivity toward Phe inhibition of ADT activity. Sequences of ADTs from algae to flowering plants were analyzed. The number of species (Number of sp.) and the number of total ADT sequences identified (Total seq.) are indicated for each taxon. The conservation of the residues listed, or a chemically similar residue, that characterize type-I ADTs and determine high sensitivity toward Phe as negative effector, is indicated in green. Yellow color indicates the substitution ratio of these residues by their equivalent counterparts in type-II ADTs or similar residues, promoting a relaxed regulation by Phe. Gray color indicates the occurrence of alternative residues.

The phylogenetic analysis of the ADTs from the two fern species fully sequenced up to date, Azolla filiculoides and Salvinia cucullata (Li et al., 2018), indicate that pteridophytes hold type-II isoforms (Supplemental Figure S8). Conversely, the ADTs from the three nonvascular plants with available genomes, Marchantia polymorpha, Physcomitrium patens, and Sphagnum phallax, were found to cluster outside the type-II group. A more detailed examination of the two lineages of ADT isoforms revealed that genes encoding for type-I enzymes have retained an intron–exon structure from the green algae ancestors, whereas genes encoding for type-II isoforms typically lack introns, including type-II enzymes from pteridophytes (Supplemental Figure S9). Intronless ADT coding genes are also present in M. polymorpha, P. patens, S. phallax, and S. moellendorfii, despite the lack of typical type-II enzymes. Overall, these results suggest that de-regulated ADTs emerged in the lineage of the Euphyllophytes from a gene duplication event that likely affected the common lineage of land plants, in which former intron–exon structure of the gene was lost.

Discussion

Effector-mediated regulation of enzymatic activity is a key control mechanism to maintain amino acid homeostasis (Figure 1). In the present work, we report that type-I and -II ADT isoforms from spermatophytes differ in their response to Phe as a negative effector. Type-I isoforms, which are more closely related to enzymes from algae and bacteria, exhibit tight inhibition by Phe, the product of the reaction. Previous literature reported IC₅₀ values for ADT activity at ∼35 µM of Phe in crude plant extracts (Jung et al., 1986). These reports are similar to our observations for the recombinant type-I enzyme PpADT-G (Table 1 and Figure 2). In contrast, type-II ADTs remain active at relatively high Phe levels, as shown for recombinant PpADT-C and other enzymes from various plants (Figures 2 and 3). Consequently, the overexpression of type-II enzymes in plant leaves promotes the accumulation of considerably higher levels of Phe than the overproduction of type-I enzymes (Figure 4). Interestingly, the overexpression of AtADT4, a type-II enzyme, was previously demonstrated to cause a strong impact on anthocyanin accumulation, likely as a result of reduced sensitivity to Phe inhibition (Chen et al., 2016). Phylogeny- and structure-guided mutagenesis studies have identified a set of residues, from the Phe binding region in the ACT regulatory domain of the enzyme, that is involved in the decreased inhibition observed in type-II enzymes. Phylogenetic evidence indicates that type-II ADTs emerged from type-I isoforms at some point in the evolution of tracheophytes, with a foreseeable impact on the massive production of Phe-derived compounds that takes place in this group of plants.

Our description here of a clade of ADT enzymes with relaxed feedback inhibition is in keeping with previous reports affecting other key enzymes of AAA biosynthesis in plants. In this regard, anthranilate synthase (AS) has two isoforms in flowering plants, named as constitutive and inducible, that differ in their sensitivity toward Trp as inhibitor. Constitutive AS, which is expressed at basal levels in different plant tissues, has a K_i for Trp in the range from 2 to 3 μM. On the other hand, the inducible isoform of AS, which is expressed in response to certain stimuli that presumably involve the synthesis of major amounts of Trp, is notably less sensitive to Trp as negative effector (K_i from 100 to 300 μM; Bohlmann et al., 1996; Song et al., 1998). Chorismate mutase (CM), the commitment step enzyme that channels chorismate into the biosynthesis of Phe and Tyr, is inhibited by both amino acids, whereas Trp promotes its activation (reviewed by Maeda and Dudareva, 2012). The inhibition of plastidial CM by Phe seems to be highly divergent between different groups of plants: K_i has been estimated to be 1.1 mM in Papaver somniferum (Benesova and Bode, 1992) and 550 µM in Solanum tuberosum (Kuroki and Conn, 1988), whereas the reported IC₅₀ values were 50 µM in A. thaliana (Westfall et al., 2014), 82 µM in Amborella trichopoda (Kroll et al., 2017), and 2.6 and 7.4 mM for the two isoforms found in P. patens (Kroll et al., 2017). Nevertheless, flowering plants possess a cytosolic isoform of CM, which is insensitive to feedback regulation by AAAs (Eberhard et al., 1996; Westfall et al., 2014). Relative to Tyr biosynthesis, it has been shown that plants in the order Caryophyllales have developed an ADH isoform with relaxed feedback inhibition by Tyr, in a close evolutionary relationship with the production of Tyr-derived betalain pigments in many species from this order (López-Nieves et al., 2018). In a very recent report, Yokoyama et al. (2021) have shown that chorismate and caffeate inhibit all three Arabidopsis 3-deoxy-d-arabino-heptulosonate 7-phosphate synthases, while Agn has the ability to counteract the chorismate-mediated inhibition of isoforms DHS1 and DHS3, but not DHS2. The occurrence of these deregulated enzymes highlights that, in many cases, the production of specialized metabolites in vascular plants has evolved to provide a surplus of precursors from the primary metabolism.

Previous works have addressed the kinetic characterization of recombinant ADTs from A. thaliana, petunia, and maritime pine without testing putative Phe feedback inhibition (Cho et al, 2007; Maeda et al., 2010;El-Azaz et al., 2016). Up to date, this aspect has only been analyzed for three recombinant ADTs: the Oyza sativa bifunctional ADT/PDT enzyme (OsPDT) (Yamada et al., 2008), and the Arabidopsis thaliana ADT2 isoform (AtADT2) (Huang et al., 2010; Chen et al., 2016), and Arabidopsis thaliana ADT4 isoform (AtADT4) (Chen et al., 2016). In the present work, the kinetic characterization of PpADT-G and PpADT-C demonstrates major differences in the mechanism underlying feedback inhibition by Phe. In PpADT-G, we observed that Phe produces a decrease in both K_m and V_max, an effect that is characteristic of uncompetitive inhibition. This model assumes that the inhibitor is only able to bind the enzyme–substrate complex but not the free enzyme. On the other hand, we have shown that PpADT-C is only scarcely affected by the presence of Phe. Although the precise inhibitory mechanism of PpADT-C remains unclear, our data (Supplemental Figure S2C) show that the inhibitor produces a decrease in the apparent affinity for the substrate and limited impact on V_max, only detectable in the presence of high Phe (>200 µM), standing in clear contrast with an uncompetitive inhibition model (in which both K_m and V_max are decreased). These differences in the mechanism of feedback-inhibition of both types of ADTs could be of major physiological relevance. Limited impact of Phe on the V_max of PpADT-C could facilitate a recovery in activity upon Agn accumulation, even in the presence of high concentration of Phe. This mechanism would not be possible under an uncompetitive inhibition model as observed for PpADT-G. These mechanistical differences may have a strong impact over Phe levels in planta, as observed in Figure 4. The in silico modeling of the enzymes (Figure 6) suggest that deregulation of type-II ADTs is of multifactorial origin: a decreased affinity toward Phe, changes in the flexibility of the peptidic backbone in the PAC region that would impact dynamics of the transition between the relaxed and the tense forms, and a different conformation and stability of the allosteric cavity in the tense form due to steric hindrance. Factors affecting the architecture and dynamics of the allosteric cavity could underlie the different mechanism of inhibition observed between PpADT-G and PpADT-C. On the other hand, the interchange of the PAC region between PpADT-G and PpADT-C improves not only PpADT-C sensitivity to Phe, but also the apparent affinity for Ppa (El-Azaz et al., 2016). It suggests, along with enzyme’s folding prediction, that PpADT-C has less active site plasticity and is in general a more rigid molecule than PpADT-G. Type-I PAC sequence could confer a more flexible and dynamic R conformation, increasing active site plasticity for the use alternative substrates (Agn or Ppa). Additionally, L320F mutation, that increases Phe IC₅₀ of the PDT reaction provides first evidence that the tetramer, and not only the dimer, could be important to achieve the tense conformation of the enzyme (Tan et al 2008). Although we have shown that the PAC region, as a whole, is a determinant element on the sensitivity to Phe inhibition of ADT and PDT reactions, we have also observed that the substitution of individual residues within this region can give a very different output depending on the substrate used (Agn or Ppa). This observation points to a complex regulatory mechanism integrating multiple elements from both the regulatory and catalytic domains. Although it was not tested, it may be hypothesized that other mutant versions of PpADT-G with increased sensitivity to Phe inhibition of PDT reaction (mutations affecting residues Ala³¹⁴, Ala³¹⁶, and Val³¹⁷) could, in contrast, deregulate ADT reaction.

An in-depth evolutionary study of the ADT family indicates that deregulated ADTs from type-II are widespread in seed plants, but absent in green algae, bryophytes, and liverworts (Figure 7; Supplemental Figure S4). Our study indicates that pteridophytes, the sister group of seed plants (Li et al., 2018), have unequivocally type-II ADTs. The analysis of 425 sequences from ferns revealed that the key substitutions T303A, L304H/Q, P308T, and L320F, which contribute to modulate the Phe-mediated inhibition of the enzyme, have similar frequencies to those found in seed plants (Figure 7;Supplemental Dataset S1). These residues usually co-occur in the same enzyme, typically encoded by genes without introns in both pteridophytes and spermatophytes. Lycophytes, the most primitive of the extant vascular plants, constitute a key group to track the emergence of type-II isoforms. In these plants, isoforms with T303A were observed frequently, as well as A314V, A316S, and V317A, being encoded by intron-less genes (Supplemental Figure S9). These last residues, although being a characteristic of type-II ADTs, were shown to decrease the IC₅₀ of PDT reaction in the model enzyme PpADT-G (Supplemental Figure S5). The remaining mutations that define type-II enzymes, especially those predicted to have important structural impact according to our model (Figure 6), were found to be rare in lycophytes. It is unlikely that ADTs from this group are deregulated isoforms. Based on the evidence provided, we propose that the primitive condition of ADTs in the Viridiplantae lineage was a high sensitivity to feedback inhibition by Phe, as observed in type-I isoforms. The structure of the ADT gene family across land plants indicates that this family suffered an early duplication event that affected all embryophytes, accompanied by the loss of introns. These duplicates were retained in vascular plants, and diverged into type-II ADTs at some point after the separation of modern lycophytes and pteridophytes, probably in a stepwise accumulation of key mutations affecting regulatory properties.

The reduction of the allosteric control of Phe biosynthesis has obvious consequences for vascular plants. As metabolism of phenylpropanoids, and particularly lignin biosynthesis, emerged and diversified, the demand of Phe for supplying such downstream pathways dramatically increased. A feasible hypothesis would be that type-I ADTs were not able to properly respond to the increasing demand for Phe in the incipient vascular plants, as far as these enzymes are inhibited when Phe accumulates at relatively low levels. This limitation seems particularly striking when we consider that the bulk of Phe biosynthesis takes place within a confined subcellular compartment, the plastids (Jung et al., 1986; Rippert et al., 2009). As the biosynthesis of phenylpropanoids occurs in the cytosol, Phe must be exported toward this compartment to be further metabolized, a process that has been found to be a major limiting factor in lignin biosynthesis rates (Guo et al., 2018). Hence, an ADT enzyme efficiently inhibited at low levels of Phe would remain mostly inactive in the limited space of plastid stroma, where Phe cannot be readily transformed into downstream products, and exportation across the plastid membrane is limiting (Widhalm et al., 2015). Along the evolution of land plants, the inconvenience of a tight feedback inhibition of ADT activity would have favored the stepwise accumulation of mutations in previously duplicated ADT genes, reducing sensitivity to Phe as negative effector. Partially deregulated ADT isoforms would overpass the restrictive allosteric control of the pathway making them more amenable for sustaining high Phe biosynthesis rates able to fueling a range of evolved pathways. This hypothesis is supported by previous studies in diverse plant species, indicating that type-II enzymes have a highlighted role in the biosynthesis of Phe-derived compounds. However, the disruption of the type-II AtADT4 and AtADT5 in A. thaliana has the largest impact on reducing lignin accumulation (Corea et al., 2012). In this regard, in silico analysis using ATTED-II version10.1 (Obayashi et al., 2018) shows that the expression of genes encoding AtADT4 and AtADT5 are strongly correlated with genes directly involved in lignin biosynthesis, including two isoforms of Phe-ammonia lyase, three isoforms of 4-coumarate-CoA ligase, the single isoforms of cinnamate 4-hydroxylase and caffeoyl CoA 3-O-methyltransferase and a lignin associated cinnamyl alcohol dehydrogenase isoform. Additionally, these genes are strongly correlated with AtMYB63 (Zhou et al., 2009), a transcriptional activator of the lignin biosynthetic pathway (Supplemental Figure S10). Moreover, available gene co-expression networks for Arabidopsis’ ADT (AtADT1-5) in ATTED-II (https://atted.jp/) show that “biosynthesis of phenylpropanoids” is on the top of the correlated pathways for genes encoding type-II ADTs while it is absent for type-I genes (Supplemental Figure S11). Other studies have also shown that the overexpression of AtADT4 resulted in Phe hyperaccumulation and elevated levels of anthocyanins, an effect that was not observed when type-I enzymes were overexpressed (Chen et al., 2016). Likewise, Petunia x hybrida ADT1 has been identified as the major contributor to Phe-derived volatile emission in the petals (Maeda et al., 2010). Finally, in Pinus pinaster, we have recently reported that the expression of PpADT-A, encoding a type-II ADT, is induced in response to the transcriptional reprograming that leads to the formation of compression wood, a specialized vascular tissue enriched in lignin (El-Azaz et al., 2020).

Notably, the appearance of type-II ADTs in spermatophytes and pteridophytes was accompanied by the retention, in all the analyzed species, of at least a type-I enzyme. Although future investigations are necessary, we hypothesize that type-I enzymes have been retained during evolution of euphyllophytes as housekeeping enzymes synthesizing Phe that is required to maintain general cell functions as required to maintain general cell functions as protein biosynthesis or alternative pathways that are only needed of basal levels of Phe.

In conclusion, our results point to the progressive development of a clade of ADT isoforms in vascular plants with relaxed feedback inhibition by Phe. Reduced regulation of the ADT activity and Phe production must have a great impact on the biosynthesis of lignin and other phenylpropanoids. Moreover, the identification of sequence motifs responsible for this trait provides an interesting biotechnological target that could help to rationally engineer the production of AAAs and their derived compounds.

Materials and methods

DNA constructs

The cloning of the coding region from PpADT-A, PpADT-C, PpADT-G, AtADT1, and AtADT2 was described previously (El-Azaz et al., 2016, 2018). Additional ADTs from P. trichocarpa (Potri11G4700, Potri4G188100), N. benthamiana (Niben8991). Arabidopsis thaliana (AtADT4), C. sativus (Cucsa52640), O. sativa (Orysa4G33390), and Z. mays (Zeama2G125923) were cloned from cDNA of the corresponding plants (see Supplemental Table S2 for primer list). All constructs for protein heterologous expression in Escherichia coli were cloned into pET30b using the NdeI/NotI restriction sites. Putative plastid transit peptide was removed and C-terminal poly-His tag was added. Mutant chimeric proteins PpADT-Gmut¹ and PpADT-Cmut¹ were generated by fusion PCR as described previously (El-Azaz et al., 2016). PpADT-Gmut⁶², Gmut¹⁰², and Gmut¹⁰³ were generated by site-directed mutagenesis using the construct pET30b-PpADT-G as template (see Supplemental Table S2 for primers). The generation of the remaining mutant versions of PpADT-G by site-directed mutagenesis was described in El-Azaz et al. (2016). Plant expression constructs were cloned into the Gateway vector pDONR™207 and recombined into pGWB11 (CaMV P35S promoter, c-terminal FLAG tag; courtesy of Dr Tsuyoshi Nakagawa, Department of Molecular and Functional Genomics, Shimane University, Japan).

Protein production and purification

Recombinant ADTs were expressed in E. coli strain BL21 DE3 RIL. After optic density at 600 nm reached 0.5–0.6, cultures were chilled on ice for 10 min before adding IPTG to a final concentration of 0.5 mM. Induced cultures were incubated for 18–20 h at 12°C under gentle shaking (∼75 rpm). Pellets were collected by centrifugation and preserved frozen at −20°C. Poly-His-tagged recombinant proteins were purified using a nickel resin (Protino Ni-TED 2000 Packed Columns, Macherey-Nagel) and exchanged to Tris buffer 50 mM pH 8.0 in Sephadex G-25 M resin (PD-10 Columns; GE Healthcare, Chicago, IL, USA).

Enzyme assays

PDT activity was determined based on the method described by Fischer and Jensen (1987a). Enzyme assays were performed in Tris buffer 50 mM pH 8.0 with around 5 µg of purified recombinant protein in a final reaction volume of 50 µL. After incubation, the reactions were stopped with the addition of 950 µL of NaOH 2M, mixed immediately, and left to settle for at least 10 min at room temperature. Absorbance was registered at 321 nm. Authentic phenylpyruvate was used to do a calibration curve at this wavelength. Prior to enzyme assays, Ppa content in the commercial preparation was estimated by treatment with HCl 1N during 15 min at room temperature, to produce its spontaneous decarboxylation to phenylpyruvate, followed by the quantification of this last compound as described before.

ADT activity was determined in a reaction mixture consisting of ∼1 µg of purified recombinant protein in Tris buffer 50 mM pH 8.0, to a final volume of 50 µL. Agn was prepared as described by El-Azaz et al. (2018) based on the previous protocol published by Rippert and Matringe (2002). Reactions were stopped with 20 µL of methanol, centrifuged at top speed for 5 min and filtered through a 0.22-µm nylon filter. Phe production was determined by ultra performance liquid chromatography/mass spectrometry (UPLC/MS) using a Waters Acquitiy UPLC system coupled to a triple quadrupole detector (Waters Milford, MA, USA). About 1 µL of the sample was subjected to separation using an Acquity UPLC BEH C18 column (1.7 µm; 2.1 × 50 mm) at a flow rate of 0.3 mL/min at 5°C, in the following gradient: 1 min in 0.12% acetic acid (v/v) in water, 3 min in a linear gradient from the later solution to methanol 100%, and 1 min in methanol 100%. Column was rinsed for 2 min with 0.12% acetic acid (v/v) in water between samples. The column effluent was analyzed by positive electrospray ionization under the following settings: capillary voltage 2.45 kV, cone voltage 10 V, source temperature 150°C, and desolvation temperature 400°C. Identification and quantification of Phe were based on a calibration curve of pure Phe.

All reactions were incubated at 30°C in gentle agitation. Enzymes were preincubated for 5 min before starting the reaction with the addition of the substrate (Ppa or Agn). When the assay was performed in the presence of Phe as negative effector, this amino acid was added to the initial reaction mixture and preincubated with the enzyme for 5 min. In all cases, measurements were done at least in duplicate, and the linearity of the reaction was corroborated for a minimum of 5 min after starting the reaction. All kinetic adjustments and parameters were obtained by using the Solver tool included in Microsoft Excel.

Overexpression of ADTs in planta

Full-length coding regions were cloned into the Gateway destination vector pGWB11 (courtesy of Dr Tsuyoshi Nakagawa, Department of Molecular and Functional Genomics, Shimane University, Japan) following standard procedures. Overexpression constructs were transformed into Agrobacterium tumefaciens strain C58C1. Saturated cultures were adjusted to an optic density at 600 nm of 0.25, and were combined with an equal density of a culture carrying the p19 construct (Voinnet et al., 2003). The 5-week-old N. benthamiana plant were infiltrated, approximately between 2 and 4 h before the end of the light period, with four different bacterial clones per plant distributed into the halves of two fully expanded leaves. Infiltration pattern was rationally designed to equally distribute 16 replicates for each overexpression construct between the plants included in the experiment. Samples were collected 72 h after infiltration, and frozen immediately in liquid nitrogen. Leave midrib and major veins were excluded from the sampling to avoid distortion of the dry weight.

Metabolite extraction

Plant samples were grinded in liquid nitrogen and combined in eight pairs of replicates for each transgenic protein. Around 100 mg of frozen powder was lyophilized at −40° C during 48 h. De-hydrated powder was resuspended in 400 µM of extraction buffer (2-amino-2-methyl-1-propanol 0.5% (v/v) pH 10.0 in EtOH 75% (v/v); Qian et al., 2019) and incubated overnight in the cold under vigorous shaking. After incubation, tubes were centrifuged for 5 min at 20,000g and 300 µL were recuperated from the supernatant, vacuum dried, dissolved in 100 µL of purified water, and frozen at −80° C until analysis. High-grade commercial Phe (20 nmol) was added as internal standard for estimating recovery rate in the control samples, which was around 75%. Phe levels were determined by UPLC/MS as described in the previous section for ADT activity. Phe content was multiplied by 4/3 to correct the estimated recovery rate, and normalized to the dry weight of the sample.

Protein extraction and western blot analysis

Total proteins from plants were extracted in buffer A (Tris buffer 50 mM pH 8.0, glycerol 10% (v/v), sodium dodecyl dulfate (SDS) 1% (w/v), EDTA 2 mM, and β-mercaptoethanol 0.1% (v/v)). Around 100 mg of frozen powder was resuspended in 200 μL of buffer A at room temperature. Samples were centrifuged at 20,000g for 5 min and 80 μL was recovered from the supernatant, mixed with 20 μL of 5X Laemmli buffer and denatured at 100°C for 5 min. The remaining supernatant was subjected to removal of excess SDS following the method described by Zaman and Verwilghen (1979), prior to determination of the protein concentration using a commercial Bradford reagent (Bio-Rad, Hercules, CA, USA; protein assay).

For immunodetection of transiently expressed ADTs in the protein extracts, 30 μg of total proteins was separated by sodium dodecyl dulfate polyacrylamide gel electrophoresis (SDS-PAGE). Western blot analysis was used following standard procedures. Transgenic proteins were detected taking advantage of the FLAG tag included in the construct, using a specific commercial antibody (OctA-Probe mouse monoclonal antibody, Santa Cruz Biotechnology) at 1:500 dilution.

Phylogenetic analysis

Phylogenetic analyses were performed in MEGA X (Kumar et al., 2018) using the Maximum-Likelihood method. The confidence probability is expressed from 0 to 100 and was estimated using the bootstrap test (1,000 replicates). The evolutionary distances were computed using the Poisson correction method (Zuckerkandl and Pauling, 1965) and are in the units of the number of amino acid substitutions per site. All positions containing gaps and missing data were eliminated.

Homology molecular modeling and model analysis

PpADT-G and PpADT-C 3D models with and without L-Phe bound to ACT regulatory domain were obtained by homology modeling using the software package Modeller 9v11 (Eswar et al., 2006; http://salilab.org/modeller). X-ray crystallographic structures from Staphylococcus aureus and Chlorobium tepidum PDTs (PDB codes 2qmw and 2qmx) were used as template for the active and L-Phe inhibited states of the enzyme, respectively (Tan et al., 2008). Protein sequence alignment of PpADT-G and PpADT-C with each template used for the homology modeling (see Supplemental Material) was constructed on the basis of the alignment of all the available P. pinaster ADTs isoforms (from -A to -I; El-Azaz et al., 2016) and both bacterial templates using Clustal Omega (Sievers and Higgins, 2014). Ten models were generated for each protein. Discrete Optimized Protein Energy score was used to select the bests models (Shen and Sali, 2006), which were submitted to Molprobity (Chen et al., 2010; http://molprobity.biochem.duke.edu/) for additional verification of their stereochemical quality. PyMOL was used for visualization of the structures, mutation simulation, and imaging generation. Morph simulations of the conformational changes between active (R) and inhibited (T) forms upon L-Phe binding were realized using Chimera (Pettersen et al., 2004). R and T 3D models for PpADT-G and C (respectively, Supplemental Figures S12–S15) and Morph simulations (Supplemental Figures S16 and S17) for the conformational changes are available as downloadable files in supplementary material in pdb format.

Co-expression analysis in A. thaliana

Co-expression analyses were conducted using the A. thaliana prediction database ATTED-II version 11.0 (http://atted.jp). AtADT4 and AtADT5 genes were used as query genes. The analysis is based on 12,686 Arabidopsis Affymetrix ATH1 GeneChip microarray data and 14,741 RNA-seq runs available from https://atted.jp/download/. The threshold was based on the Logit Score as measure (>3). Top correlated pathways to AtADT1-5 based on the kyoto encyclopedia of genes and genomes (KEGG) annotation were directly obtained from ATTED-II version 11.0.

Accession numbers

Amborella trichopoda protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): Ambtr8033 (AMTR_s00080p00093010), Ambtr37103 (AMTR_s00037p00204590), Ambtr00040233 (AMTR_s00040p00217260), and Ambtr3137 (AMTR_s00003p00164060).

Aquilegia coerulea protein sequences were obtained from Genbank: Aqcoe4G202500 (PIA24954.1), Aqcoe1G230700 (PIA57807.1), Aqcoe1G110700 (PIA54757.1), and Aqcoe7G047700 (PIA49648.1).

Arabidopsis thaliana genes involved in this article are listed under the following accession numbers as in TAIR (www.arabidopsis.org/): AtADT1 (At1g11790), AtADT2 (At3g07630), AtADT3 (At2g27820), AtADT4 (At3g44720), AtADT5 (At5g22630), and AtADT6 (At1g08250).

Azolla filiculoides protein sequences were obtained from the Fernbase (https://www.fernbase.org/): Azofi 0149g053188 (Azfi_s0149.g053188), Azofi0096g043700 (Azfi_s0096.g043700), Azofi0074g037290 (Azfi_s0074.g037290), Azofi0071g036867 (Azfi_s0071.g036867), and Azofi 233g059377 (Azfi_s0233.g059377).

Brachypodium_distachyon protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): Bradi1G16517 (BRADI_1g16517v3), Bradi1G65800 (BRADI_1g65800v3), Bradi4G38380 (BRADI_4g38380v3), Bradi5G09030 (BRADI_5g09030v3), Bradi5G09020 (BRADI_5g09020v3), and Bradi3G31590 (BRADI_3g31590v3).

Cucumis sativus protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): Cucsa052640 (Csa_6G289730), Cucsa200040 (Csa_1G145980), Cucsa385350 (Csa_6G151110), Cucsa045180 (Csa_6G513690), Cucsa072650 (Csa_2G417830), and Cucsa200050 (Csa_1G145970).

Eucalyptus grandis protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): EucgrJ00958 (EUGRSUZ_J00958), EucgrE04043 (EUGRSUZ_E04043), EucgrA01119 (EUGRSUZ_A01119), and EucgrJ00428 (EUGRSUZ_J00428).

Glycine max protein sequences were obtained from Phytozome v13 (https://phytozome-next.jgi.doe.gov/): GlymaU021400 (Glyma.U021400.1.p), Glyma13G319000 (Glyma.13G319000.1.p), Glyma12G181800 (Glyma.12G181800.1.p), Glyma12G085500 (Glyma.12G085500.1.p), Glyma11G189100 (Glyma.11G189100.1.p), Glyma09G004200 (Glyma.09G004200.1.p), Glyma12G193000 (Glyma.12G193000.1.p), and Glyma17G012600 (Glyma.17G012600.1.p).

Marchantia polymorpha protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): Marpo0024s0086 (MARPO_0024s0086), Marpo0024s0087 (MARPO_0024s0087), Marpo0071s0057 (MARPO_0071s0057), and Marpo0204s0012 (MARPO_0204s0012).

Nicotiana benthamiana protein sequences were obtained from The Sol Genomics Network (https://solgenomics.net/): Niben8991 (Niben101Scf08991g02002.1) and Niben5382 (Niben101Scf05382g00001.1).

O. sativa protein sequences were obtained from The Rice Genome Annotation Project (http://rice.uga.edu/): Os07g49390.1, Os03g17730.1, Os04g33390.1, Os09g39230.1, Os09g39260.1, and Os10g37980.1.

Physcomitrium patens protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): Phypa3c1722810 (Pp3c17_22810), Phypa3c184480 (Pp3c18_4480), Phypa3c29380 (Pp3c2_9380), and Phypa3c1913030 (Pp3c19_13030).

Picea abies protein sequences were obtained from the ConGenie database (https://congenie.org/): Picab7947 (MA_7947g0030), Picab339590 (MA_339590g0010), Picab647307 (MA_647307g0010), Picab14041 (MA_14041g0010), Picab82198 (MA_82198g0010), Picab825097 (MA_825097g0010), and Picab10433230 (MA_10433230g0010).

Pinus pinaster protein sequences were obtained from GenBank: PpADT-A (APA32582.1), PpADT-B (APA32583.1), PpADT-C (APA32584.1), PpADT-D (APA32585.1), PpADT-E (APA32586.1), PpADT-F (APA32587.1), PpADT-G (APA32588.1) PpADT-H (APA32589.1), and PpADT-I (APA32590.1).

Petunia hybrida protein sequences were obtained from GenBank: PethyADT1 (FJ790412), PethyADT2 (FJ790413), and PethyADT3 (FJ790414).

Populus trichocarpa protein sequences were obtained from the version 3 PopGenIE (https://popgenie.org/): Potri11G4700 (Potri.011G004700), Potri8G195500 (Potri.008G195500), Potri4G013400 (Potri.004G013400), Potri4G188100 (Potri.004G188100), and Potri9G148800 (Potri.009G148800).

Salvinia cucullata protein sequences were obtained from the Fernbase (https://www.fernbase.org/): Salcu0212g025961 (Sacu_v1.1_s0212.g025961), Salcu0039g012193 (Sacu_v1.1_s0039.g012193), Salcu0162g023967 (Sacu_v1.1_s0162.g023967), Salcu0005g002563 (Sacu_v1.1_s0005.g002563), Salcu0029g010023 (Sacu_v1.1_s0029.g010023), Salcu0029g010076 (Sacu_v1.1_s0029.g010076), and Salcu0010g004784 (Sacu_v1.1_s0010.g004784).

Selaginella moellendorfii protein sequences were obtained from EnsemblPlants (http://plants.ensembl.org/): Selmo118675 (SELMODRAFT_118675), Selmo95583 (SELMODRAFT_95583), and Selmo142600 (SELMODRAFT_142600).

Spagnum fallax protein sequences were obtained from Phytozome v13 (https://phytozome-next.jgi.doe.gov/): Sphfa0226s0033 (Sphfalx03G027000.1.p), Sphfa0137s0058 (Sphfalx06G126300.1.p), Sphfa0001s0251 (Sphfalx07G103900.1.p), and Sphfa0021s0031 (Sphfalx13G095600.4.p).

Zea mays protein sequences were obtained from Phytozome version 13 (https://phytozome-next.jgi.doe.gov/): Zeama2G141273 (Zm00001eb013110), Zeama2G145451 (Zm00001eb397170), Zeama2G437912 (Zm00001eb084260), Zeama2G125923 (Zm00001eb421790), Zeama2G466543 (Zm00001eb096110), and Zeama2G115841 (Zm00001eb404610).

Staphylococcus aureus PDT (PDB code 2qmw).

Chlorobium tepidum PDT (PDB code 2qmx).

Supplemental data

The following materials are available in the online version of this article.

Supplemental Table S1. Determination of the kinetic parameters of PpADT-G and PpADT-C for the ADT reaction.

Supplemental Table S2. Primers used in the generation of pET30b constructs.

Supplemental Figure S1. Direct representation of v0 against Agn concentration for PpADT-G and PpADT-C.

Supplemental Figure S2. Determination of enzyme activity at different Phe concentrations.

Supplemental Figure S3. Western blot of plant extracts from N. benthamiana leaves over-expressing type-I and -II ADTs.

Supplemental Figure S4. Alignment of the ACT domain.

Supplemental Figure S5. Activity of PpADT-Gmut⁹, Gmut¹¹, Gmut¹², Gmut¹³, and Gmut¹⁶ in response to Phe concentration.

Supplemental Figure S6. Sequence alignment of PpADT-G and C (without the predicted plastid transit peptide) and ADT/PDTs from S. aureus and C. tepidum.

Supplemental Figure S7. Co-occurrence of selected amino acids involved in the feedback inhibition of type-I ADTs from spermatophytes.

Supplemental Figure S8. Phylogenetic analysis of ADTs from nonseed plants, showing the existence of type-II in pteridophytes.

Supplemental Figure S9. Intron–exon structure in the ADT genes in the freshwater algae Klebsormidium nitens, the nonvascular plants M. polymorpha and P. patens, the lycophyte S. moellendorfii, the pteridophyte A. filiculoides and the flowering plant A. thaliana, and the corresponding sequence of the PAC domain.

Supplemental Figure S10. Co-expression analysis of AtADT4 and AtADT5.

Supplemental Figure S11. Top correlated pathways to Arabidopsis ADT encoding genes.

Supplemental Figure S12. PpADT-C_relaxed (PDB file).

Supplemental Figure S13. PpADT-C_tense (PDB file).

Supplemental Figure S14. PpADT-G_relaxed (PDB file).

Supplemental Figure S15. PpADT-G_tense (PDB file).

Supplemental Figure S16. PpADT-C-R-to-T (PDB file).

Supplemental Figure S17. PpADT-G-R-to-T (PDB file).

Supplemental Dataset S1. Co-occurrence of amino acids at the Phe binding cavity of plant ADTs.

 

J.E.A. designed the research, performed research, and wrote the paper with contributions of all authors; F.M.C. designed the research, project leader; B.B. performed computational analysis; C.A. designed the research, project leader; F.T. designed the research and performed research.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/general-instructions is Fernando de la Torre (fdelatorre@uma.es).