The entry reaction of the plant shikimate pathway is subjected to highly complex metabolite-mediated regulation.

The plant shikimate pathway directs bulk carbon flow toward biosynthesis of aromatic amino acids (AAAs, i.e. tyrosine, phenylalanine, and tryptophan) and numerous aromatic phytochemicals. The microbial shikimate pathway is feedback inhibited by AAAs at the first enzyme, 3-deoxy-d-arabino-heptulosonate 7-phosphate synthase (DHS). However, AAAs generally do not inhibit DHS activities from plant extracts and how plants regulate the shikimate pathway remains elusive. Here, we characterized recombinant Arabidopsis thaliana DHSs (AthDHSs) and found that tyrosine and tryptophan inhibit AthDHS2, but not AthDHS1 or AthDHS3. Mixing AthDHS2 with AthDHS1 or 3 attenuated its inhibition. The AAA and phenylpropanoid pathway intermediates chorismate and caffeate, respectively, strongly inhibited all AthDHSs, while the arogenate intermediate counteracted the AthDHS1 or 3 inhibition by chorismate. AAAs inhibited DHS activity in young seedlings, where AthDHS2 is highly expressed, but not in mature leaves, where AthDHS1 is predominantly expressed. Arabidopsis dhs1 and dhs3 knockout mutants were hypersensitive to tyrosine and tryptophan, respectively, while dhs2 was resistant to tyrosine-mediated growth inhibition. dhs1 and dhs3 also had reduced anthocyanin accumulation under high light stress. These findings reveal the highly complex regulation of the entry reaction of the plant shikimate pathway and lay the foundation for efforts to control the production of AAAs and diverse aromatic natural products in plants.


INTRODUCTION
The shikimate pathway directs carbon flow from central carbon metabolism to the biosynthesis of aromatic amino acids (AAAs)-L-tyrosine, L-phenylalanine and Ltryptophan (Tyr, Phe and Trp, respectively)-and numerous aromatic natural products.
Since AAAs are required for protein synthesis in all organisms but animals lack the shikimate pathway, AAAs are essential human nutrients and the shikimate pathway is the target of the most widely-used herbicide, glyphosate (Shah et al., 1986;Klee et al., 1987;Pollegioni et al., 2011;Hildebrandt et al., 2015, Figure 1). Plant natural products derived from the shikimate and AAA pathways play critical roles in plant physiology and adaptation and are widely used as nutraceuticals, pharmaceuticals, and biomaterials (Herrmann, 1995;Tzin and Galili 2010;Maeda and Dudareva, 2012, Figure 1).
Shikimate pathway intermediates are used to synthesize hydrolysable tannins and chlorogenic acids, but also to produce the anti-influenza virus agent Tamiflu (Niggeweg et al., 2004;Barbehenn and Peter Constabel, 2011;Ghosh et al., 2012). Trp is a precursor of auxin phytohormones, plant defense compounds (e.g. indole alkaloids and glucosinolates), and cancer drugs such as glucobrassicin and vincristine produced in Brassicaceae and Catharanthus roseus, respectively (De Luca and St Pierre, 2000;Kliebenstein, 2011;Saini et al., 2013;Sanchez-Pujante et al., 2017). Phenylpropanoids derived from Phe are the largest class of plant natural products, next to terpenoids, and include flavonoids, anthocyanin pigments, tannins, the principal cell wall component lignin, etc. (Boerjan et al., 2010;Vogt, 2010). As lignin can account for up to 30% of deposited carbon in vascular plants (Razal et al., 1996;Boerjan et al., 2010), plants direct a significant portion of carbon flow through the shikimate pathway. Despite the paramount and broad importance of these aromatic plant natural products in both plants and humans, little is known about how the shikimate pathway is regulated in plants, which represents a major knowledge gap in plant biology and biochemistry.
The shikimate pathway converts phosphoenolpyruvate (PEP) and D-erythrose 4-Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 found in plants and some bacteria such as Mycobacterium tuberculosis (Gosset et al., 2001;Richards et al., 2006;Tohge et al., 2013). Although their reaction mechanisms are similar (e.g. metal requirements; Walker et al., 1996;Shumilin et al., 1999;Hartmann et al., 2003;Webby et al., 2010), type II DHSs appear to have more complex regulatory mechanisms than type I enzymes. Unlike type I DHSs having a single allosteric effector binding site (Shumilin et al., 1999;Hartmann et al., 2003), M. tuberculosis type II DHS has at least two effector binding sites for Phe and Trp, and are inhibited by the combination of Phe and Trp, but not by individual ones (Webby et al., 2010).
The biochemical properties of AthDHS2 and AthDHS3 and the metabolite-mediated feedback regulation of Arabidopsis DHS enzymes remain to be explored.
To understand the regulatory mechanisms of the committed step of the shikimate pathway, here we generated recombinant enzymes of all three DHS isoforms of Arabidopsis thaliana and conducted their detailed biochemical characterization. All three AthDHSs had similar kinetic parameters; however, Tyr and Trp specifically inhibited AthDHS2, but not AthDHS1 and AthDHS3. We further identified several other pathway intermediates including chorismate and caffeate, that strongly inhibit all AthDHS isoforms. DHS activity of Arabidopsis and spinach leaf extracts were not inhibited by AAAs, consistent with prior reports (Huisman and Kosuge, 1974;Pinto et al., 1986;Sharma et al., 1993), but this was due to lower expression of AthDHS2 and the loss of the AthDHS2 AAA sensitivity in the presence of AAA-insensitive AthDHS1.
Analyses of Arabidopsis dhs knockout mutants further showed some distinct roles of AthDHS isoforms in planta. Together, these biochemical and genetic data revealed the highly complex metabolite-mediated regulatory mechanisms of the entry step of plant shikimate and AAA pathways, catalyzed by three distinct DHS enzymes.

Biochemical characterization of three Arabidopsis DHS enzymes
To biochemically characterize plant DHSs, all three Arabidopsis DHS enzymes, AthDHS1, AthDHS2, and AthDHS3, were expressed as hexa-histidine tagged recombinant proteins in E. coli and purified using affinity chromatography. To test metal and reducing agent requirements, DHS assays were conducted in the presence and absence of DTT and different metal ions at pH 7. All three AthDHSs showed the highest activity with Mn 2+ and DTT (Figure 2A and Supplemental Figure 2), consistent with the prior report for AthDHS1 (Entus et al., 2002). No DHS activity was detectable in AthDHS1 and AthDHS3 without DTT, whereas some residual DTTindependent activity was detected in AthDHS2 (Figure 2A). Similar results were obtained at pH 8 (Supplemental Figure 2). Other divalent cations, Co 2+ , Cd 2+ ,Cu 2+ and Zn 2+ , used by M. tuberculosis DHS (Webby et al., 2005), were also tested, but only Cd 2+ partially supported DHS activity in all three AthDHSs at 20 to 50% of corresponding Mn 2+ -dependent activity (Supplemental Figure 2). Co 2+ -dependent DHS activity, previously detected from plant tissue extracts (Rubin and Jensen, 1985;Ganson et al., 1986;Morris et al., 1989;Doong et al., 1992), was not observed in any of the AthDHS recombinant enzymes (Supplemental Figure 2). DHS assays conducted at different pH from 6.8 to 8.0 showed the maximum activity at pH 7.4 for all three AthDHSs ( Figure 2B), consistent with the previous report on DHS activity from maize, potato (Solanum tuberosum) and spinach tissue extracts (Graziana and Boudet, 1980;Pinto et al., 1986;Doong et al., 1993). Thus, all three AthDHSs require a reducing agent, prefer Mn 2+ as a metal cofactor, and have an optimal pH of 7.4.
To check potential contamination of E. coli DHS enzymes in our enzyme preparation, DHS assays were conducted using E. coli cells carrying the empty or AthDHS1 vector in the presence of Mn 2+ or Fe 2+ , as Fe 2+ supports the activity of E. coli DHS (Stephens and Bauerle, 1991) but not AthDHS1 (Entus et al., 2002). Unlike in their crude supernatant, no Fe 2+ -dependent DHS activity was detected in either sample after purification, whereas only the purified AthDHS1 sample showed Mn 2+ -dependent activity (Supplemental Figure 3), confirming that E. coli DHS activity was effectively removed to undetectable levels.
Steady-state enzyme kinetic analyses were then conducted using various Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 concentrations of E4P and PEP substrates ( Table 1 and Supplemental Figure 4). All reactions followed Michaelis-Menten kinetics, with the exception of AthDHS2, which was inhibited at high E4P concentrations beyond 4 mM (Supplemental Figure 4C). All three AthDHSs showed much higher K m toward E4P (1.6 to 2.8 mM) than PEP (250 to 706 µM, Table 1), consistent with previous reports in spinach and carrot (Doong et al., 1992;Suzuki et al., 1996). Overall catalytic efficiency (k cat /K m ) was the highest for AthDHS1 followed by AthDHS2 and then AthDHS3 (62, 25, and 12 s -1 mM -1 , respectively, for PEP), which reflect their turnover rates (k cat , Table 1). Although AthDHS1 is the most efficient enzyme, all three AthDHSs overall showed similar ranges of K m and k cat .

Tyrosine and tryptophan inhibit Arabidopsis DHS2, but not DHS1 and 3.
To test if plant DHS enzymes are regulated by AAAs, like microbial DHSs (Bentley, 1990), recombinant AthDHS activity was monitored in the presence of AAAs. Since Mycobacterium tuberculosis DHS, a close microbial homolog of plant DHSs, requires both Trp and Phe for its inhibition, we first used the mixture of three AAAs (Tyr, Phe, and Trp) at 1 mM. The activity of AthDHS1 and AthDHS3 was slightly activated by the AAA mixture up to 120%, whereas AthDHS2 activity was significantly inhibited (P<0.001, Figure 2C). When individual AAAs were tested, AthDHS1 was significantly activated by Tyr and Trp, whereas only Trp activated AthDHS3 ( Figure 2D). In contrast, 60 to 70% of AthDHS2 activity was inhibited by Tyr or Trp, but not by Phe ( Figure 2E). The combination of Tyr and Trp had no additive inhibitory effect on AthDHS2 activity. The addition of Phe had no major effect on the AthDHS2 inhibition by Tyr or Trp ( Figure 2E). AthDHS2 assays with varying concentrations of individual AAAs ranging from 1 µM to at least 5 mM further revealed that Tyr and Trp, but not Phe, inhibit AthDHS2 with the IC 50 values of 230.4 and 225.1 µM, respectively ( Figure   2F). Phe slightly activated the AthDHS2 activity at a very high concentration but only beyond 5 mM ( Figure 2F). The other seventeen proteinogenic amino acids did not significantly alter activities of any AthDHS isoforms (Supplemental Figure 5) (Figure 1). Shikimate, prephenate, and arogenate at 1 mM did not significantly affect the activity of AthDHS1 and AthDHS3, while AAAs again slightly activated them ( Figure 3A). On the other hand, the activity of AthDHS2 was inhibited by shikimate, prephenate, and arogenate by approximately 30, 25 and 75%, respectively ( Figure 3A). Notably, all three AthDHSs were completely inhibited by chorismate at 1 mM ( Figure 3A). DHS assays at varied concentrations of chorismate ranging from 1 µM to 1 mM showed that AthDHS1, AthDHS2, and AthDHS3 are inhibited by chorismate at the IC 50 values of 97.3, 52.5 and 83.0 µM, respectively ( Figure 3B).
Since commercially-available chorismate reagents include impurities, we further evaluated if chorismate is indeed the inhibitor that reduces DHS activity. Before adding to the DHS assays, the chorismate solution was incubated with the active or boiled CM2 enzyme of Arabidopsis (AthCM2), which specifically converts chorismate into prephenate (Westfall et al., 2014). Untreated chorismate completely inhibited AthDHS1 activity as expected, but chorismate that was incubated with active AthCM2 did not (Supplemental Figure 6A). Chorismate that was treated with boiled AthCM2 exhibited the same inhibitory effect as one with the untreated chorismate (Supplemental Figure   6A). These results confirm that chorismate indeed inhibits AthDHS activity.

AthDHS3.
Next, we examined if other effector molecules potentially exert additive or synergistic effects on the chorismate-mediated inhibition of AthDHSs. When the AAA mixture was combined with chorismate, chorismate-inhibited AthDHS1 and AthDHS3 activity recovered slightly to one fifth and half, respectively, of their corresponding activity without any effector molecules ( Figure 3C). Notably, when arogenate was provided with chorismate (both at 1 mM), chorismate-inhibited AthDHS1 and AthDHS3 activity Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 fully recovered to the levels equivalent to no-effector controls ( Figure 3C). The chorismate-inhibited AthDHS2 activity was not recovered by the combination of chorismate and arogenate ( Figure 3C and D), likely because arogenate by itself also inhibits AthDHS2 (Figure 3A). The other shikimate pathway intermediates, shikimate and prephenate, did not attenuate chorismate-dependent inhibition of AthDHS1 (Supplemental Figure 7). An assay using various concentrations of arogenate in the presence of 1 mM chorismate revealed that arogenate offsets the chorismate-mediated inhibition of AthDHS1 and AthDHS3 activity with the IC 50 values of 343.9 and 305.4 µM, respectively ( Figure 3D).
Since arogenate is not commercially available, we prepared the arogenate reagent through transamination of prephenate with aspartate (see Methods;Maeda et al., 2010;Schenck et al., 2015), which may still be present in the arogenate preparation and contribute to the above observed effect. However, unlike arogenate, 1 mM prephenate or aspartate did not affect the DHS activity regardless of the presence of chorismate (Supplemental Figure 7A). Also, before adding it to the assays, we incubated arogenate with hydrochloric acid (HCl), which converts arogenate into Phe (Gilchrist and Connelly, 1987). The AthDHS1 reaction containing the HCl-treated arogenate and chorismate did not offset chorismate-mediated DHS inhibition and still showed no enzymatic activity, like that with only chorismate (Supplemental Figure 6B). These results together revealed that arogenate counteracts the inhibition of AthDHS by chorismate.

AthDHS enzymes.
Since Phe itself did not significantly affect activity of all AthDHS isoforms ( Figure 2D and E), the question still remains: how do DHS enzymes monitor the pathway activity of the Phe branch of AAA biosynthesis? To address this question, we tested five intermediate compounds in the downstream phenylpropanoid pathway for their effects on AthDHS activity ( Figure 4A). Although cinnamate, p-coumarate, ferulate, and sinapate had no effects on all AthDHSs at 1 mM, caffeate completely inhibited activities of all three AthDHSs ( Figure 4B). Caffeoyl shikimate, one of the major derivatives of caffeate in plants (Boerjan et al., 2010;Vogt, 2010), also fully inhibited Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 AthDHS1 and 2 and reduced AthDHS3 activity by 70% ( Figure 4C). Although pcoumarate by itself had no effect, p-coumaroyl shikimate partially reduced the activity of all AthDHSs by 66 to 75% ( Figure 4C). DHS assays of individual DHS isoforms with varying concentrations of caffeate further showed that AthDHS1, 2, and 3 are inhibited by caffeate with IC 50 values of 49.7, 69.2 and 53.4 µM for AthDHS1, 2, and 3, respectively ( Figure 4D). Thus, the phenylpropanoid intermediates, caffeate and its derivative, effectively inhibit all three DHS enzymes of Arabidopsis.
Expression pattern and ratio of AAA-inhibited vs. non-inhibited DHS isoforms determine AAA sensitivity of DHS activity detected from plant tissues.
To evaluate if the results of the recombinant DHS enzymes can also be observed in plant tissue-derived DHS activity, total protein extracts were prepared from fully expanded mature leaves of Arabidopsis and spinach and subjected to DHS assays using different inhibitors. Chorismate and caffeate at 1 mM reduced the total DHS activity in both Arabidopsis and spinach leaf extracts by more than half ( Figure 5A and B), consistent with their inhibitory effects on all three AthDHSs (Figure 3 and 4). The AAA mixture at 1 mM, by contrast, had no significant effects on the DHS activity of both Arabidopsis and spinach extracts (Figure 5A and B). This observation is consistent with previous reports (Huisman and Kosuge, 1974;Pinto et al., 1986;Sharma et al., 1993) but contradicts the results of the recombinant AthDHS2 that was inhibited by the same AAA treatment ( Figure 2C).
Initially, we hypothesized that the AthDHS2 gene expression may be very low compared to that of AthDHS1 and AthDHS3, which encode DHS enzymes that are not inhibited by AAAs ( Figure 2C). To test this possibility, the copy numbers of different AthDHSs were compared by reverse transcription quantitative PCR (RT-qPCR) in mature leaves of Arabidopsis that were harvested at the same stage as for the above DHS activity assays. However, the level of the AthDHS2 transcripts were still >50% of AthDHS1 and roughly 7-fold more abundant than AthDHS3 ( Figure 5C). In roots, AthDHS1 and AthDHS2 were expressed at similar levels, which are much higher than those of AthDHS3 ( Figure 5C). These results indicate that the AAA-inhibited AthDHS2 is still expressed at substantial levels in mature tissues, even though overall DHS activity does not exhibit any inhibition by AAAs (Figure 5A and B). Thus, an Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 additional factor must be contributing to the lack of the observed AAA inhibition of DHS activity (Figure 5A and B).
Next, we thought that the presence of non-inhibited DHS isoforms (e.g. AthDHS1) may affect the AAA sensitivity of AthDHS2, given that AthDHSs are known to function as tetrameric or dimeric forms (Suzich et al., 1985;Pinto et al., 1986;Webby et al., 2005). To test this, the recombinant enzymes of AAA non-inhibited AthDHS1 and AAA-inhibited AthDHS2 were mixed with different ratios (0, 0.25, 0.5, 0.75, and 1) and DHS activity assays were conducted with different inhibitors. Based on the kinetic parameters and the regulatory behaviors of individual AthDHSs, the expected level of DHS activity was first calculated and plotted (dotted lines in Figure   5D, see Methods). Without any inhibitors or with Phe, the observed DHS activity in the various AthDHS1 and AthDHS2 mixtures matched with the theoretical plot (black and purple lines, respectively, in Figure 5D), consistent with the absence of inhibitory effects of Phe on both AthDHS1 and AthDHS2 (Figure 2D and E). Notably, in the presence of 1 mM Tyr and Trp, however, observed DHS activity was significantly higher than theoretically calculated activities in any of the AthDHS1 and AthDHS2 mixtures (orange and magenta lines, respectively in Figure 5D). Similar results were obtained for the AthDHS2 and AthDHS3 mixture ( Figure 5E). These results suggest that the presence of AthDHS1 and AthDHS3 reduces the sensitivity of AthDHS2 to AAAs, which likely contributes to the observed lack of AAA-mediated inhibition of DHS activity detected from the leaf extracts ( Figure 5A and B).
Gene Ontology (GO) analyses of publicly available co-expression data suggest that AthDHS2 is co-expressed with genes involved in plastid development and photosynthesis, whereas AthDHS1 and AthDHS3 are associated with other shikimate, AAA, and phenylpropanoid genes (Supplemental Figure 8 and Supplemental Table   1; Peltier et al., 2004;Ytterberg et al., 2006;Obayashi et al., 2018). Unlike AthDHS1 and AthDHS3, which are strongly expressed in response to pathogens and elicitors, AthDHS2 is often induced upon changes in light conditions, based on expression databases (Supplemental Figure 9; Hruz et al., 2008;Klepikova et al., 2016).
Transcriptome data from eFP browser suggests that AthDHS2 tends to express predominantly in early developmental stages ( Figure 5F and Supplemental Figure   9C). Also, DTT-independent DHS activity was detected only in AthDHS2 but not in Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 AthDHS1 or AthDHS3 (Figure 2A). Thus, we thought that AthDHS2 may be highly expressed in young developing seedlings with low photosynthetic activity (and thus limited reducing energy; Winter et al., 2007). Our RT-qPCR analysis showed up to 10fold higher expression of AthDHS2 than AthDHS1 or AthDHS3 in 3 to 4-day-old etiolated and de-etiolated seedlings of Arabidopsis ( Figure 5G).
With the predominant expression of AthDHS2 in photosynthetically less-active tissues, such as etiolated and de-etiolated seedlings, we then rationalized that AAAmediated DHS inhibition may be detectable. Since Arabidopsis seedlings are too small to obtain enough enzyme extract for the DHS assay, we tested this hypothesis using deetiolated whole seedlings of spinach, whose genome also contains an AthDHS2-like ortholog (Supplemental Figure 1). DHS activity from the spinach seedling extracts was significantly reduced in the presence of AAAs as well as chorismate and caffeate ( Figure 5H). Thus, plant tissue-derived DHS activity can be also inhibited by AAAs when AAA-sensitive AthDHS2 is predominantly expressed such as in young seedlings; however, when AthDHS2 expression is not predominant, such as in mature leaves, the AAA-sensitivity of the AthDHS2 activity is not observable not only due to its low expression ( Figure 5C) but also because the presence of AthDHS1 and AthDHS3 makes AthDHS2 insensitive to AAA (Figure 5D and E).
Arabidopsis dhs1 and dhs3 mutants are hypersensitive to Tyr and Trp, respectively.
To further investigate in planta functions of the AthDHS isoforms, we obtained and characterized T-DNA insertional mutants of AthDHS1, AthDHS2, and AthDHS3 (dhs1, dhs2, and dhs3, respectively) from Arabidopsis thaliana. A T-DNA fragment was present in the 3rd and 1st exons of the dhs1 and dhs3 mutants, respectively, and in the 3rd intron of dhs2 ( Figure 6A). Their corresponding transcripts were not detectable by RT-PCR and RT-qPCR analyses in these mutants (Figure 6B and C). AthDHS1 and AthDHS3 gene expression was upregulated 1.5-to 2-fold in the dhs2 mutant, but no significant change of AthDHS transcript levels were detected in dhs1 or dhs3 ( Figure   6C). No visual growth phenotype was observed for any of the dhs mutants under standard conditions used in this study ( Figure 6D). Given the observed isoform-specific regulation of the AthDHS2 enzyme by AAAs (Figure 2), the effects of AAA treatment on these dhs mutants were examined by growing them on growth media containing different concentrations of individual AAA.
All of the dhs mutants grown without any AAAs were again indistinguishable from Col-0 wild type and had similar root lengths (Supplemental Figure 10). High concentrations of Phe beyond 300 µM suppressed the root growth of all dhs mutants but similarly to Col-0, with the IC 50 values of ~550 µM (Supplemental Figure 10B).  Figure   10B). Also, none of the dhs mutants showed sensitivity to up to 1000 µM of shikimate (Supplemental Figure 10B).
To test if the observed Tyr and Trp sensitivity of the dhs1 and dhs3 mutant, respectively, are indeed due to the loss of AthDHS1 and AthDHS3, their wild-type CDS genes were introduced to the dhs1 and dhs3 mutant backgrounds using their native promoters (see Materials and Methods). The dhs1 and dhs3 mutants carrying the respective AthDHS1 and AthDHS3 rescue construct (dhs1 AthDHS1 and dhs3 AthDHS3) recovered the root growth phenotype under high Tyr and Trp conditions, respectively (Supplemental Figure 11). The dhs1 AthDHS1 and dhs3 AthDHS3 lines behave very similarly to Col-0. The introduction of the empty vector into the dhs1 and dhs3 mutants (dhs1 Empty and dhs3 Empty) did not result in growth recovery under high AAA conditions (Supplemental Figure 11). These results further support that the lack of AthDHS1 and AthDHS3, but not an unknown secondary mutation(s), resulted in the hypersensitivity to Tyr and Trp in dhs1 and dhs3, respectively.
To better understand the inhibitory effects of Tyr and Trp, we additionally supplied other AAAs and/or shikimate to the media containing Tyr or Trp at 300 µM.
The addition of shikimate at 500 µM had no effect on the Tyr-and Trp-mediated growth inhibition of dhs1 and dhs3, respectively (Supplemental Figure 12). Addition of 300 µM Trp with Tyr, however, resulted in similar root length between dhs1 and Col-0 ( Figure 7A and B). Further addition of Phe, together with Trp, drastically recovered the stunted root growth of dhs1 in the presence of Tyr, although Phe alone was not sufficient to exert such effect (Figure 7A and B). The Trp sensitivity of dhs3 was not recovered by the addition of Tyr or Phe individually; however, when both Tyr and Phe were added together, dhs3 showed drastic recovery in root growth and exhibited the root length similar to Col-0 ( Figure 7C and D). The observed recovery of Tyr and Trp sensitivity of dhs1 and dhs3 by Phe together with Trp and Tyr, respectively, suggest that the depletion of other AAAs resulted in the inhibition of their root growth.  Figure 14A). Thus, the lack of AthDHS2, the only DHS isoform inhibited by Tyr (Figure 2), prevents the Tyr sensitivity of Arabidopsis thaliana plants.

Tyr sensitivity of Arabidopsis plants is prevented by lack of the Tyr-inhibited
We then hypothesized that the Tyr treatment restricts carbon flow through the shikimate pathway in Arabidopsis, but not in the dhs2 mutant that lacks Tyr-inhibited AthDHS2. To test this hypothesis, Col-0 and dhs mutants were treated with and without Tyr (at the same 300 µM as above) and/or glyphosate-an herbicide that inhibits 5enolpyruvylshikimate-3-phosphate synthase (EPSPS) enzymes in planta (Figure 1) and hence promotes shikimate accumulation, which likely reflects the difference in carbon flow through the shikimate pathway (Hollander and Amrhein, 1980;Klee et al., 1987;Pollegioni et al., 2011). The glyphosate treatment without Tyr indeed increased shikimate levels as expected, but similarly between the genotypes (Figure 8B, left).
Feeding Tyr at 300 µM by itself (without glyphosate treatment, Figure 8B, right) did not alter the shikimate level in all genotypes, except for the elevated shikimate content Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 of dhs1, expressed in µmol/g FW, due to its extremely small shoot size ( Figure 8A).
After the treatment with glyphosate, the dhs2 plants grown with the exogenous Tyr showed 5-fold higher accumulation of shikimate, while other genotypes exhibited only 2-to 3-fold increase (Figure 8B, right). Introducing the wild-type AthDHS2 gene in the dhs2 mutant almost completely eliminated the glyphosate-induced elevation of shikimate under high Tyr condition (Supplemental Figure 14B). These results suggest that the Tyr treatment restricts shikimate production through the Tyr-mediated negative feedback inhibition of AthDHS2 enzymes in Arabidopsis plants.
To further test if the absence of AthDHS1 or AthDHS3 in dhs1 or dhs3 leads to AAA-mediated inhibition of DHS activity, which was not observed in Col-0 ( Figure   5A), DHS activity was analyzed from the crude extracts of 4-week-old leaves of Col-0 and dhs mutants in the presence of individual AAAs. However, DHS activity was not inhibited by any AAAs at 1 mM in any of the dhs mutants (Supplemental Figure 15). This is likely because the remaining AthDHS3 and AthDHS1 in dhs1 and dhs3, respectively, is sufficient to mask the AthDHS2-mediated feedback inhibition of DHS activity by AAAs in the crude extract.

High light-induced phenylpropanoid production is attenuated in dhs mutants.
To further characterize in vivo functions of AthDHSs, the dhs mutants and Col-0 plants were subjected to metabolite analyses using gas chromatography-mass spectrometry and liquid chromatography-mass spectrometry (GC-MS and LC-MS, respectively), first using leaf tissues of four-week-old plants grown under standard growth conditions. Compared to Col-0 the dhs1 mutant accumulated less Phe, Asp, and Glu, while the dhs3 mutant exhibited less Phe and Ala (Supplemental Table 2A). The levels of the other amino acids including Tyr and Trp were not significantly different among all of the dhs mutants and Col-0 (Supplemental Table 2A). The levels of AAA-derived specialized metabolites such as glucosinolates, flavonols and tocopherols were comparable between Col-0 and the dhs mutants (Supplemental Table 2A). These results suggest that the lack of individual AthDHS genes has minor impacts on overall plant phenotypes ( Figure 6D) and metabolite levels (Supplemental Table 2A), with the exception of dhs1 and dhs3-specific alteration in some amino acid levels. Given that various stress conditions induce the production of AAA-derived natural products (Bohinc et al., 2012;Landi et al., 2015;Liu and Lu, 2016) and the expression of some DHS genes (Pinto et al., 1988;Dyer et al., 1989;Keith et al., 1991;Devoto et al., 2005;Yan et al., 2007), these dhs mutants were subjected to stress However, no significant and consistent differences were observed between Col-0 and dhs1 in the accumulation of these metabolites (Supplemental Figure 16).

High intensity light (HL) stress induces biosynthesis of Tyr-derived tocopherols
and Phe-derived phenylpropanoid compounds, such as flavonols and anthocyanin pigments (Das et al., 2011;Landi et al., 2015). To investigate the roles of different AthDHS isoforms in elevated production of Tyr and Phe-derived metabolites, Col-0 and the dhs mutants were grown at 100 µE for 4 weeks and then subjected to 650 µE of HL treatment ( Figure 9A). Overall the levels of AAAs and many of their derivatives were elevated after 2 days of the HL treatment (Supplemental Table 2B), with the exception of reduced Phe levels likely due to its rapid utilization for phenylpropanoid biosynthesis (Tohge et al., 2013;Chen et al., 2016). AAA levels were overall similar among genotypes after the 2-day HL treatment, except for slightly higher Tyr in dhs1 than other genotypes (Supplemental Table 2B).
The HL treatment rapidly induced Phe-derived anthocyanin pigments, which became visible after 2 days in the abaxial surfaces of leaves ( Figure 9A). However, this was less pronounced in dhs mutants, with dhs1 and dhs3 having significantly less anthocyanin levels than Col-0 after 2 days ( Figure 9B). This metabolic phenotype was repeatedly observed at 2 and 5 days of the HL treatment (Supplemental Figure 17) and was rescued by introducing the AthDHS gene into each corresponding dhs mutant (Supplemental Figure 18). Flavonol quercetin glycosides, such as quercetin-3-Orhamnoside-7-O-rhamnoside (Q3R7R), were also induced strongly after the HL treatment, but accumulated significantly less in all three dhs mutants than in Col-0 ( Figure 9C and Supplemental Table 2B). The levels of other flavonols, kaempferol glycosides, and a hydroxycinnamate, sinapoyl-O-glucoside, were also elevated but less Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 pronounced than quercetin derivatives or anthocyanins, and were not significantly different among genotypes ( Figure 9D and E and Supplemental Table 2B). The levels of another hydroxycinnamate, sinapoyl-malate, were slightly but significantly decreased after 2-day HL treatment, and showed no significant differences between genotypes ( Figure 9E right and Supplemental Table 2B). The levels of glucosinolates were overall similar among genotypes, except one aliphatic glucosinolate, 7methylsulphinylheptyl-glucosinolate (7MTH), which was higher in dhs1 and dhs3 than Col-0 (Supplemental Table 2B). The HL treatment increased the levels of Tyr-derived lipophilic antioxidants, alpha-and gamma-tocopherols, but similarly among all genotypes (Supplemental Table 2B). These metabolic phenotypes of the dhs mutants demonstrates that AthDHSs play important roles in the elevated production of phenylpropanoid compounds, such as quercetin derivatives and anthocyanin pigments, under HL stress.

Plant DHS enzymes exhibit high K m towards E4P and have varied redox dependency.
Unlike in microbes, limited information is available on the biochemical properties of plant DHS enzymes, which catalyze the entry step for biosynthesis of AAAs and numerous plant natural products (Figure 1). Kinetic analyses of the recombinant AthDHS enzymes showed approximately 10-fold higher K m values for E4P than those for PEP (Table 1), consistent with activity data for partially purified DHS from spinach leaves (Doong et al., 1992). The K m values of AthDHSs for PEP were similar to those of microbial DHSs that have equivalent K m for E4P and PEP ( Table 1; Liu et al., 2008;Reichau et al., 2016). On the other hand, the K m of AthDHSs for E4P is 10-fold higher than those of microbial DHSs. Although detection of intracellular levels of E4P has not been successful from plant tissues (Arrivault et al., 2009), the in vivo E4P concentration is likely much higher in plant cells than in non-photosynthetic organisms due to the presence of the Calvin-Benson cycle (the reductive pentose phosphate pathway), which provides an additional source of E4P besides the oxidative pentose phosphate pathway. Thus, the availability of E4P likely has a significant impact on overall activity of DHS and hence the shikimate pathway in Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 planta. A prior study showed that the suppression of transketolase, which is responsible for the E4P production in the pentose phosphate pathways, led to a major reduction in AAAs and phenylpropanoid compounds in tobacco (Nicotiana tabacum) plants (Henkes et al., 2001). Thus, an enhanced supply of E4P, such as by upregulating the pentose phosphate pathways, likely plays an important role in maintaining high carbon flux through the shikimate pathway in plants.
All three AthDHSs were found to be redox-dependent (Figure 2A), which was also observed in a prior study for AthDHS1 (Entus et al., 2002). Although the underlying mechanism remains to be determined, the redox regulation of plant DHS enzymes can contribute to the functional coupling of the pentose phosphate and shikimate pathways (Mousdale and Coggins, 1985;Ganson et al., 1986;Maeda and Dudareva, 2012). Notably, however, residual but substantial levels of redoxindependent DHS activity was detected, but only in AthDHS2 (Figure 2A), which belongs to a distinct phylogenetic clade from AthDHS1 and AthDHS3 (Supplemental Figure 1). Actively growing young tissues, for example, have developing plastids with limited photosynthetic activity, but have a high demand for AAAs, along with other amino acids, to support rapid growth (Pyke, 1999; Jarvis and López-Juez, 2013; Hildebrandt et al., 2015). Under such conditions, the unique redox-independent AthDHS2 activity may allow Arabidopsis plants to maintain the basal levels of the shikimate pathway activity without redox activation ( Figure 10B).

The DHS2 enzyme is inhibited by Tyr and Trp in Arabidopsis.
Microbial DHS enzymes are directly feedback inhibited by AAAs, which tightly controls the entry step of AAA biosynthesis to maintain cellular amino acid and metabolic homeostasis (Bentley 1990). By contrast, previous studies found that plant DHS activities are not inhibited by AAAs (Huisman and Kosuge, 1974;Pinto et al., 1986;Sharma et al., 1993); this has long puzzled plant biochemists seeking to understand how the shikimate pathway is regulated in plants (Herrmann, 1995;Tzin and Galili, 2010;Maeda and Dudareva, 2012;Lynch and Dudareva, 2020). Our initial assays of mature leaf tissues of Arabidopsis and spinach also failed to detect AAA inhibition ( Figure 5A and B). Surprisingly, however, we found that Tyr and Trp inhibit recombinant AthDHS2, but not AthDHS1 and AthDHS3 (Figure 2), which was Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 observed many times in independent experiments (e.g. Figure 3A and C).
Further analyses revealed at least two factors that may explain why AAAmediated inhibition of DHS activity was rarely detected from plant tissue extracts. First, the expression levels of AAA-inhibited AthDHS2 was the same as or lower than that of AAA non-inhibited AthDHS1 in mature leaf and root tissues (Figure 5C), where the majority of DHS activity has been analyzed (Huisman and Kosuge, 1974;Pinto et al., 1986;Sharma et al., 1993). Second, the AAA sensitivity of AthDHS2 was attenuated when AthDHS2 was mixed with AthDHS1 or AthDHS3 (Figure 5D and E), which are not inhibited by AAAs (Figure 2C and D). A similar case of one enzyme affecting a property of another isoform was also previously found in poplar (Populus trichocarpa) 4-coumaric acid:CoA ligase 5 (4CL5) in monolignol biosynthesis that alters substrate specificity of another isoform 4CL3 through heterocomplex formation (Chen et al., 2014). Although the molecular mechanism behind this intriguing observation requires further investigation, when AAA non-inhibited AthDHS1 is predominantly expressed, such as in mature leaves, AthDHS1 masks the AAA-mediated inhibition of AthDHS2 ( Figure 10A). By contrast, in young seedlings where AthDHS2 was predominantly expressed, we can detect AAA-mediated inhibition of DHS activity ( Figure 10B).
This study further showed that the dhs1 and dhs3 mutants are sensitive to Tyr and Trp, respectively (Figure 7 and Supplemental Figure 10), the former was also noted in a prior thesis study (Crowley, 2006). Further, we were able to fully rescue these phenotypes by the introduction of the corresponding wild-type AthDHS genes (Supplemental Figure 11). Although we do not know why dhs1 and dhs3 are specifically sensitive to Tyr and Trp, respectively, these findings provide in vivo evidence that the lack of AthDHS1 or AthDHS3 increases AthDHS2 sensitivity to Tyr and Trp inhibition, respectively, in planta. Conversely, the dhs2 mutant was resistant to Tyr (Figure 8).
Prior and current studies showed that Arabidopsis is sensitive to AAAs, especially to high Tyr concentration (Voll et al., 2004;de Oliveira et al., 2019), and exhibits a cupshaped leaf phenotype ( Figure 8A, Supplemental Figure 13). A similar cup-shaped phenotype was observed when the feedback-insensitive prephenate or arogenate dehydratase was expressed in Arabidopsis plants, elevating the levels of Phe (Tzin et al., 2009;Huang et al., 2010) . The dhs2 mutant, however, did not show the cup-shaped  (Hollander and Amrhein, 1980;Pollegioni et al., 2011), led to significantly elevated shikimate accumulation in Tyr-treated dhs2 compared to Col-0 ( Figure 8B). These in vitro and in vivo data together demonstrate that the AthDHS2 enzyme is indeed inhibited by Tyr and Trp in Arabidopsis plants.

Roles of different AthDHS enzymes in Arabidopsis.
Why do AAAs inhibit one DHS isoform (i.e., AthDHS2) but not the others in plants?
Similar to microbes, in young seedlings, where AthDHS2 is predominantly expressed  Figure 10B). By contrast, the IC 50 (or K i ) values of downstream TyrA and AS enzymes are ~4 to 10-fold lower than those of AthDHS2 (Li and Last, 1996;Rippert and Matringe, 2002b;Kanno et al., 2004;Schenck et al., 2015;Schenck et al., 2017;Lopez-Nieves et al., 2018). This suggests that when cellular concentrations of Tyr and Trp are increased in planta, TyrA and AS activities will be initially inhibited before AthDHS2. The feedback regulation of AthDHS2 by Tyr and Trp may then act as a second layer of regulation to ensure that excess carbon will not flow into the shikimate pathway. High levels of free amino acids can accumulate transiently during developmental transitions, though difficult to detect, and are observed during senescence, when proteins are actively degraded (Soudry et al., 2005;Hirota et al., 2018) and AthDHS2 is highly expressed (Supplemental Figure 9). Thus, the regulation of AthDHS2 likely allows tight control of overall AAA levels and metabolic homeostasis under certain developmental conditions such as early seedling growth and Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 senescence.
In contrast to AthDHS2, AthDHS1 and AthDHS3 were not inhibited and were even slightly activated by AAAs (Figure 2). AthDHS1 was the major DHS gene to be expressed in roots and mature leaves of Arabidopsis ( Figure 5C). AthDHS1 and AthDHS3 are co-expressed with genes involved in AAA biosynthesis as well as AAAderived compounds, such as Trp and Phe-derived specialized metabolites, respectively (Supplemental Figure 8 and Supplemental Table 1, Obayashi et al., 2018). Also, AthDHS1 and AthDHS3 are strongly induced upon various biotic and abiotic stresses based on Arabidopsis expression databases (Supplemental Figure 9), which is consistent with orthologs of AthDHS1 and AthDHS3, but not of AthDHS2, from different plants that are also responsive to various stresses (Pinto et al., 1988;Dyer et al., 1989;Keith et al., 1991;Devoto et al., 2005;Yan et al., 2007). While the induction of Trp and Phe-derived compounds were largely unaltered in dhs1 upon MeJA treatment at different concentrations (Supplemental Figure 16), HL-induced accumulation of Phederived anthocyanin pigments was substantially reduced in dhs1 and dhs3 and to a lesser extent in dhs2 ( Figure 9B and Supplemental Figure 17). HL-induced production of flavonol quercetin derivatives were also significantly lower in all dhs mutants than Col-0 ( Figure 9C). Other phenylpropanoids, such as kaempferol and sinapoyl derivatives, were not significantly different between genotypes, likely due to their limited induction after the HL stress ( Figure 9D and E). Therefore, in mature leaves, especially under stresses, AthDHS1 and AthDHS3 likely allow rapid induction of total DHS activity without being inhibited by AAAs and by masking the AthDHS2 sensitivity to AAAs, together leading to efficient induction of phenylpropanoid compounds derived from Phe ( Figure 10A). Taken together, these data show that the three DHS isoforms have some distinct roles based on their mutant phenotypes, expression profiles, and distinct biochemical properties; however, these DHS enzymes also have overlapping roles and their combinatorial effects likely fine tune the regulation of the key entry step of this shikimate and AAA pathways under different conditions. Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 The shikimate/AAA pathways are regulated by multiple metabolite-mediated feedback regulatory mechanisms in Arabidopsis.
Amino acid biosynthetic pathways are typically feedback inhibited by the pathway endproducts (i.e., amino acids, Galili et al., 2016). This is the case in the post-chorismate AAA pathways in both plants and microbes (Figure 1; Bentley, 1990;Tzin and Galili, 2010;Maeda and Dudareva, 2012) and in the upstream shikimate pathway in microbes (Bentley, 1990;Shumilin et al., 1999;Hartmann et al., 2003;Webby et al., 2010). This study, however, revealed that pathway intermediates are also potentially involved in regulating the plant shikimate and AAA pathways, at least in Arabidopsis (Figure 10): Chorismate is a strong inhibitor of all three AthDHSs (with IC 50 of ~50-100 µM, Importantly, chorismate and arogenate are located at the branch points into each AAA biosynthesis pathway (Figure 1). Since two of the chorismate-utilizing enzymes, AS and CM, are feedback inhibited by Trp and Phe or Tyr, respectively, in both plants (Romero et al., 1995a;Romero et al., 1995b) and microbes (Bentley, 1990), elevated levels of all AAAs will lead to chorismate accumulation and thus inhibition of all DHS enzymes. This may not be the case when individual AAAs (e.g. only Trp) accumulate, because the carbon flow can be redirected to the other AAA biosynthesis without accumulating chorismate. Therefore, the "sequential inhibitory" mechanism, as previously proposed (Doong et al., 1993), may allow plants to accumulate substantial levels of individual AAAs without completely inhibiting the initial step of the shikimate pathway (Figure 10). Such complex regulation of DHSs, together with their transcriptional regulation (Pinto et al., 1988;Dyer et al., 1989;Keith et al., 1991;Devoto et al., 2005;Yan et al., 2007), is likely important for fine tuning the stoichiometry of AAAs and efficiently producing certain AAA-derived compounds in The effect of arogenate on different AthDHS isoforms is much more difficult to decipher. In plants, Phe and Tyr inhibit ADT and TyrA enzymes, respectively, much more efficiently (IC 50 of ~10 to 60 µM; Connelly and Conn, 1986;Siehl and Conn, 1988;Rippert and Matringe, 2002;Yamada et al., 2008) than the upstream CM enzymes (IC 50 of ~300 µM to 1 mM; Kuroki and Conn, 1988;Benesova and Bode, 1992). Therefore, theoretically speaking, arogenate accumulates only when both Phe and Tyr accumulate at the range of ~60 to 300 µM that inhibit ADT and TyrA but not CM; beyond 300 µM, CM is inhibited and arogenate will not accumulate. In tissues predominantly expressing AthDHS1 and AthDHS3 (e.g. mature leaves, Figure 5B), the accumulated arogenate offsets chorismate-mediated inhibition of AthDHS1 and AthDHS3 ( Figure 3C and D), likely allowing high chorismate accumulation and hence the production of chorismate-derived compounds such as folate, salicylic acid, indole alkaloids, and glucosinolates, (Radwanski and Last, 1995;Kliebenstein, 2011;Saini et al., 2013;Sanchez-Pujante et al., 2017;Rekhter et al., 2019). However, when Tyr and Phe accumulate further (beyond ~300 µM) and start to inhibit CM, the arogenate accumulation will be attenuated and hence chorismate will again inhibit DHSs.
Unfortunately, the cost and instability of chorismate and arogenate did not allow direct testing of their effects on DHSs and the shikimate pathway in planta, and further genetic studies are needed to decipher their regulatory functions.
Although Phe is an important inhibitor of many microbial DHS enzymes (Bentley, 1990;Hartmann et al., 2003;Webby et al., 2010), Phe had no effect on any of AthDHS activity ( Figure 2D and E) consistent with prior studies (Huisman and Kosuge, 1974;Pinto et al., 1986;Sharma et al., 1993). Since plants synthesize high levels of Phe-  (Figure 4A), the meta-hydroxyl group of caffeate is likely important for DHS inhibition, though the p-coumaroyl shikimate result suggests more complex interactions (Figure 4C). Although further in vivo analyses are needed, the current results generate an interesting hypothesis that plant DHSs may monitor the levels of downstream intermediates, such as caffeate, rather than Phe, so as to directly coordinate the regulation of the shikimate and phenylpropanoid pathways (Figure 10).

Conclusions and future perspectives
The current study revealed that plant DHS activity and hence the shikimate pathway are subjected to highly complex regulation that is mediated by multiple pathway products and intermediates (Figure 10). This is in contrast to the more straightforward regulation of microbial DHSs, which are feedback regulated by AAAs (Bentley, 1990). This radical difference between plants and microbes are likely linked to their distinct demand and usage of AAAs: in most microbes AAAs are the pathway "end products" to be mainly utilized for protein synthesis (Bentley, 1990;Shumilin et al., 1999;Hartmann et al., 2003;Webby et al., 2010), whereas plants additionally produce numerous natural products derived from the shikimate and AAA pathways (Herrmann, 1995

Plant materials
Wild-type Arabidopsis thaliana (Col-0) was grown under a 12/12 hour 100 µE light/dark cycle with 85% air humidity in soil supplied with Hoagland solution or on the agarose-containing ½-strength Murashige and Skoog (MS) medium with 1% sucrose unless stated otherwise. T-DNA insertional mutants of AthDHS1, AthDHS2 and AthDHS3 (SALK_055360, SALK_033389, and SK2559S, respectively) were obtained from the Arabidopsis Biological Resource Center (ABRC). Their homozygous T-DNA insertions were confirmed by PCR using primers listed in Supplemental Table 3.

Preparation of AthDHS protein expression vectors
For expression of AthDHS1, AthDHS2, and AthDHS3 enzymes in E. coli, the CDS fragments without sequences corresponding to their transit peptides (AthDHS1; residues 49-525, AthDHS2; residues 34-507, AthDHS3; residues 52-527) were amplified from cDNA by Phusion DNA polymerase (Thermo Fisher Scientific, Waltham, USA) using primers listed in Supplemental Table 3. The PCR products were purified using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany) and inserted into the NdeI-EcoRI sites for AthDHS1 and AthDHS3 or the NdeI-BamHI sites for AthDHS2 of the pET28a vector (Millipore Sigma, St. Louis, USA) using the In-Fusion HD cloning kit and protocol (Clontech, Mountain View, USA). All of the resulting plasmids were sequenced to confirm that no errors were introduced during PCR and cloning. The purified proteins were also analyzed by SDS-PAGE to evaluate their purity. All the purification steps were performed at 4°C unless stated otherwise.

Preparation of crude protein extract from spinach and Arabidopsis
Spinach (Spinacia oleracea) leaves were purchased at a local grocery store. For deetiolated spinach seedlings, spinach was germinated on the soil and grown in the dark for 3 days, followed by exposure to normal growth light for 1 day for de-etiolation.
More than 20 g of fully expanded mature leaves or whole de-etiolated seedlings were harvested and ground to a fine powder in a mortar and pestle with liquid nitrogen. After dissolving in 100 mL extraction buffer [20 mM HEPES (pH 7.6), 1 mM DTT and 0.1% β-mercaptoethanol] and filtrating with Miracloth, the samples were centrifuged at 10,000 × g for 10 min and subsequently at 50,000 × g for 30 min. The resulting supernatant was concentrated with Amicon Ultra Centrifugal Filters (Millipore Sigma, Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 St. Louis, USA) by centrifugation until the solution volume became less than 1 mL. The concentrated solution was desalted twice by Sephadex G50 column size-exclusion chromatography (GE Healthcare, Chicago, USA) into 1 mL of 50 mM HEPES (pH 7.4) and quantified by the Bradford assay (Bradford, 1976). All the purification steps were performed at 4°C.

Enzymatic assays
Unless otherwise noted, DHS enzymatic activity was monitored as previously described   Table 3. Four biological replicates with two technical RT-qPCR replicates were conducted. Expression of UBC21 gene (AT4G27960) was used to normalize the sample-to-sample variations between different cDNA preparations.

Gene expression analyses
In order to compare expression levels of different AthDHS isoforms, the copy numbers of individual AthDHS transcripts were determined by an absolute quantification method (Larionov et al., 2005). Briefly, the AthDHS coding regions were PCR amplified using the pET28 vectors carrying DHS sequences as the template, Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 purified using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany), and quantified using Nanodrop (Thermo Fisher Scientific, Waltham, USA). The purified PCR fragments of known concentrations were then used to generate standard curves for each gene-specific primers, which was then used to calculate actual copy numbers of each AthDHS isoform (Larionov et al., 2005). For comparison of gene expression between Col-0 and the dhs mutants, dilution series of the Col-0 cDNA was used to determine the standard curves.  13-14.5 min, 35-70%; 14.5-15.5 min, 70-99%, 15.5-17 min, 99%; 17-17.5 min, 99-10%; 17.5-20 min, 1%. The spectra were recorded using full scan mode of negative ion detection, covering a mass range from m/z 100 to 1500. The resolution was set to 25,000, and the maximum scan time was set to 250 ms. The sheath gas was set to a value of 60, while the auxiliary gas was set to 35. The transfer capillary temperature Downloaded from https://academic.oup.com/plcell/advance-article/doi/10.1093/plcell/koaa042/6067414 by guest on 23 January 2021 was set to 150°C, while the heater temperature was adjusted to 300°C. The spray voltage was fixed at 3 kV, with a capillary voltage and a skimmer voltage of 25 and 15 V, respectively. Retention times, MS spectra, and associated peak intensities were extracted from the raw files using the Xcalibur software (Thermo Fisher Scientific, Waltham, USA). For confirmation of the identity of almost all compounds, LC-MS/MS analysis was performed with normalized collision energy (NCE) 20%, observing the fragmentation patterns (Supplemental Table 4 Shikimate concentration was determined by comparing with a dilution series of the shikimate standards.

Chemical feeding experiments
Chemical feeding experiments were carried out as described previously (de Oliveira et where we grew other plants in soil. AAAs and/or shikimate were added to autoclaved media after cooling down to temperatures below 55°C. Plants were grown on the same plate side by side to minimize environmental effects, and multiple plates were prepared to obtain replications. Root lengths of 10-day-old seedlings were quantified by ImageJ.

Construction of DHS cladogram tree
DHS orthologs were first identified by BlastP searches utilizing the amino acid sequence of AthDHS1 as query against Phytozome and SpinachBase databases (Goodstein et al., 2012;Xu et al., 2017). All of the obtained sequences were then used to construct a tree of DHS genes using MEGA 7 (Kumar et al., 2016) and are available as a FASTA file in Supplemental Data Set 3. The sequences were aligned by the MUSCLE algorithm and then constructed into the tree based on the maximumlikelihood method with 1,000 bootstrap replicates (Supplemental Data Set 4).

Preparation of Chemical Compounds Used in Enzymatic Assay
For arogenate production, prephenate was enzymatically converted into arogenate by Arabidopsis prephenate aminotransferase (AT2G22250) recombinant enzyme and purified by an anion-exchange chromatography, as previously described (Maeda et al., 2010;Schenck et al., 2015).  Table 1. A list of top 20 genes co-expressed with AthDHS1, AthDHS2 and AthDHS3.
Supplemental Table 2. Levels of amino acids and AAA-derived metabolites under standard growth condition (before HL treatment) and after 2-day HL treatment in Col-0 and the dhs mutants. Table 3. Primer list used in this study. Kinetic parameters were obtained from the Michaelis-Menten curves obtained by DHS activity measured using various concentrations of the E4P or PEP substrate (Supplemental Figure 4). Since AthDHS2 exhibited substrate inhibition by E4P, the Michaelis-Menten kinetics curves were generated by assuming that those curves represent the activity that would be found if no substrate inhibition occurred (Bernstein et al., 1978). To estimate the K m and V max values, data points at high substrate concentrations were plotted according to the Lineweaver-Burk plots. Data are means ± SEM (n = 3 replicated reactions).

ACKNOWLEDGMENTS
We are     All the individual data points are shown as dots.      The shikimate pathway leads to biosynthesis of aromatic amino acids (AAAs), which are not only required for protein synthesis but also used as precursors of numerous AAA-derived natural products (green letters) in plants.      Different isoforms of plant DHS enzymes are subjected to highly-complex effector-mediated feedback regulation in a tissue-specific manner. (A) In mature leave tissues, total DHS activity is mainly governed by AthDHS1 and AthDHS3 enzymes, which are not directly inhibited by AAAs. Chorismate strongly inhibits AthDHS1 and AthDHS3, which is offset by arogenate. (B) In young etiolated/de-etiolated seedlings, the AAAinhibited AthDHS2 isoform is dominantly expressed and directly inhibited by Tyr and Trp. Chorismate and arogenate also inhibit AthDHS2. In both tissues, Phe has not regulatory effects on plant DHSs but caffeate and its derivatives act as strong inhibitors for all AthDHS isoforms, likely coordinating the shikimate and Phe phenylpropanoid biosynthesis. Blue and pink lines represent feedback inhibition of AthDHS1/3 and AthDHS2 enzymes, respectively, by indicated effectors. Gray lines with an arrowhead and hash indicate known feedback activation and inhibition, respectively, of AS, CM, TyrA and ADT enzymes. DHS, 3-deoxy-D-arabinoheptulosonate 7-phosphate synthase; AS, anthranilate synthase; CM, chorismate mutase; TyrA, arogenate dehydrogenase; ADT, arogenate dehydratase; E4P, D-erythrose 4-phosphate; PEP, phosphoenolpyruvate.