UMP inhibition and sequential firing in aspartate transcarbamoylase open ways to regulate plant growth

Pyrimidine nucleotides are essential to plant development. We proved that Arabidopsis growth can be inhibited or enhanced by down- or upregulating aspartate transcarbamoylase (ATC), the first committed enzyme for de novo biosynthesis of pyrimidines in plants. To understand the unique mechanism of feedback inhibition of this enzyme by uridine 5-monophosphate (UMP), we determined the crystal structure of the Arabidopsis ATC trimer free and bound to UMP, demonstrating that the nucleotide binds and blocks the active site. The regulatory mechanism relies on a loop exclusively conserved in plants, and a single-point mutation (F161A) turns ATC insensitive to UMP. Moreover, the structures in complex with a transition-state analog or with carbamoyl phosphate proved a mechanism in plant ATCs for sequential firing of the active sites. The disclosure of the unique regulatory and catalytic properties suggests new strategies to modulate ATC activity and to control de novo pyrimidine synthesis and plant growth.

associated subunits, meaning that both the catalytic and regulatory sites must reside within the 1 same polypeptide chain 31 . However, despite the wealth of biochemical and structural knowledge disclose unique catalytic and regulatory properties of plant ATCs, suggesting new strategies to 11 control de novo pyrimidine synthesis and plant growth. 12

15
To explore the importance of ATC for plant growth we used artificial microRNA (amiRNA) 16 to knockdown ATC in Arabidopsis. Two selected lines, atc-1 and atc-2, exhibited 16% and 10% 17 residual ATC transcript and a similar drop in protein levels compared to wild-type (WT; Col-0) 18 controls ( Fig. 1b-d). Conversely, we constitutively overexpressed ATC in two Arabidopsis lines,

25
ATC downregulated lines also showed pale leaves, suggesting lower chlorophyll levels, 26 and presumably impaired photosynthesis, whereas ATC-OX lines exhibited no phenotypic 27 alterations other than the bigger size (Fig. 1b). Because of the pale leaf coloration, four-week-old

12
we tested different N-terminal truncated forms. One construct spanning aa 82-390 (named 13 atATC) was purified as a stable homotrimer (Fig 2b), and matched in size (excluding the 2.6 kDa 14 fusion tag) the 36 kDa mature enzyme in whole leaf extracts (Fig. 2c). atATC produced diffraction 15 quality crystals readily and the structure was determined at 1.7 Å resolution (Table 1; 16 Supplementary Fig. 2). The structure of the atATC trimer resembles a three-bladed propeller with a concave face holding three active sites in between subunits, thus preserving the overall 18 architecture of the transcarbamoylase family 33 (Fig. 2d). Each subunit folds into two subdomains    26,28,34 . In addition, atATC has the CP-loop (aa 156-169) and Asp-loop (aa 309-332) (Fig. 2a,d) 24 that in other ATCs undergo large conformational movements upon substrate binding 16,28 .

25
Unexpectedly, additional electron density at the active site proved the presence of a 26 molecule of UMP captured during protein expression and kept throughout the purification and 27 crystallization process (Fig. 2d,e and Supplementary Fig. 1). The nucleotide fills the active site,

28
with the ribose in C3' endo pucker and the base in anti conformation (Fig. 2e). The phosphate 29 binds near the N-end of helix H2 and interacts with R136, T137, R187 and H215 (at the N-domain), whereas the ribose 2'-and 3'-OH bind to the side chain of R248 (C-domain). The 4-O 1 atom of the pyrimidine ring interacts with R310 (Asp-loop), and the 2-O and 3-NH bind through 2 three waters to R310, R248 and V250 (C-domain), whereas the C5 and C6 atoms make Van der 3 Waals contacts with the 349 PLP 351 loop (C-domain). In addition, the CP-loop from the adjacent 4 subunit interacts with the inhibitor through Van der Waals contacts of residues A164 and A165 5 and makes a H-bond between S162 and the phosphate (Fig. 2d,e).

6
Next, we freed the enzyme of UMP by a gel filtration procedure ( Supplementary Fig. 4) 7 and determined the crystal structure of the apo form at 3.1 Å resolution ( Table 1). The structure 8 turned to be similar to the UMP-inhibited conformation except for aa 160-166 of the CP-loop that 9 appear flexibly disordered in absence of the nucleotide (Supplementary Fig. 3).   [35][36][37] . Seedling assays in presence of 0.2 mM or 0.4 mM PALA showed respectively a decrease 15 in fresh weight to 59% or 23% of untreated seeds and a root length reduction to 29% or 6% ( Fig.   16 3a). These results support the reduced growth observed in atc downregulated lines ( Fig. 1b-d) and agree with previous PALA-inhibition studies 36 . Chlorosis was also apparent in PALA-treated 18 seedlings (Fig. 3a), further endorsing the effect of reduced ATC levels on chloroplast functionality 19 ( Fig. 1f,g).

20
To gain further insight into the reaction mechanism, we determined the structure of atATC 21 in complex with the transition-state analog at 1.6 Å resolution (Table 1). Surprisingly, the structure showed the atATC trimer with PALA bound to only one of the subunits (Fig. 3b). This subunit 23 undergoes a 10º hinge closure of the N-and C-domains and a 24º rigid body rotation of the Asp-24 loop (Fig. 3c), emulating the movement needed to bring CP and Asp in close contact to favor the 25 reaction 37,38 . In contrast, the other two subunits exhibit an open conformation, similar to the apo 26 or UMP-bound states, and have the active sites empty or with two sulfate ions and one glycerol 27 molecule from the crystallization solution (Fig. 3b). Only the CP-loop interacting with PALA is well-28 defined in the electron density map (Fig. 3b), whereas the other CP-loops are flexibly disordered.
The substoichiometric binding of PALA is remarkable, since other ATC structures proved 1 the binding of three molecules of PALA per trimer 28,[38][39][40][41]    binds to L350 (C-domain). Also, the a-carboxylate group binds to R248 (C-domain) and the b-6 carboxylate binds to R310 and Q312 (Asp-loop). In addition, the CP-loop from the adjacent 7 subunit binds through S163 to the phosphonate moiety, and places K166 at interacting distance  These results strongly suggested that despite the overall structural similarity with other 18 ATCs, the atATC trimer hides a mechanism of communication between active sites that allows 19 only one subunit to reach the closed catalytic conformation.

21
The CP-loop obstructs the simultaneous closure of the subunits

22
The explanation for the unusual binding of PALA to atATC was likely at the CP-loop, as 23 the most distinct element compared to other non-plant ATCs ( Fig. 2a and Supplementary Fig. 1).

24
This loop is flexible in absence of ligands ( Fig. 3b and Supplementary Fig. 3) but adopts two 25 distinct conformations whether UMP or PALA are bound to the adjacent subunit (Figs. 2e and 26 3d). With UMP, the CP-loop folds in an extended "inhibited" conformation, with A164, A165 and 27 S162 interacting with the nucleotide, S163 and K166 pointing outwards the active site, and the 28 side chain of F161 inserted in between subunits (Fig. 4a). In turn, upon PALA binding, the CP-29 loop rearranges into two short and nearly perpendicular 310 a-helices, placing S163 and K166 to interact with the transition-state analog and moving A164, A165 and S162 outwards the active 1 site (Fig. 4b). In this "active" conformation, F161 flips 180º compared to the position with UMP, 2 and projects towards the trimer threefold axis, where intersubunit distances are shortened by the 3 interactions between neighbor E156 residues (Fig. 4b). These tight contacts at the center of the  Two additional atATC structures reinforced this hypothesis. One structure, obtained from 7 crystals with PALA and soaked in CP (Table 1), showed a trimer with one subunit bound to PALA, 8 a second subunit with CP, and a third subunit with CP and with one glycerol and one sulfate ion 9 filling the Asp binding site (Fig. 4c). The second structure, obtained by co-crystallization with CP, 10 showed all three subunits in the trimer bound to CP (Table 1). In both structures, the three CP-

21
parallel assays with F161A proved that although the limiting rate is similar to the WT, the activity is not inhibited by UMP and becomes more sensitive to the presence of PALA (Fig. 5a). In 23 addition, F161A showed decreased activity at high substrate concentrations, whereas this 24 substrate inhibition effect was not apparent in the WT (Fig. 5b,c).

25
ITC analysis failed to detect the binding of UMP to F161A, supporting the loss of inhibition 26 by the nucleotide (Table 2). In turn, the affinity for PALA (KD PALA = 0.12 µM) is increased 5-fold compared to WT, in agreement with the enhanced inhibition, and also the affinity for CP (KD CP =0.7 conformation of the CP-loop, reducing the affinity for UMP, and thus, favoring the alternate CP-1 or PALA-bound conformation (Fig. 4a,b).

2
To better understand the effect of the mutation, we determined the structure of F161A 3 with UMP (Table 1). In apparent contradiction with the activity and ITC results, the structure 4 showed a molecule of UMP in the active site (Fig. 5d), likely favored by the high concentration of 5 nucleotide (5 mM) used during crystallization. The CP-loop is in the inhibited conformation, and 6 the missing F161 side chain is replaced by a glycerol molecule (Fig. 5e). We also determined the 7 structure of F161A crystallized with PALA (Table 1). Interestingly, the structure showed a trimer 8 bound to three molecules of PALA rather than one as in the WT (Fig. 5f). Although ITC indicated 9 that PALA binds with high affinity to only one site per trimer (Table 2)

19
However, downregulation of CPS or DHODH had no effect on potato or tobacco plant growth and 20 DHO mutants were only slightly affected 43 , whereas downregulation of UMPS led to increased 21 yield of potato tuber by overcompensation reactions from the pyrimidine salvage pathway 44 . In contrast, we showed that ATC downregulation strongly inhibits plant growth ( Fig. 1b-e), as 23 previously reported for Solanaceae and Arabidopsis plants 42,43 , causing a severe decrease in 24 photosynthetic efficiency (Fig. 1f,g). Conversely, we proved for the first time that plant growth can 25 be enhanced by ATC overexpression (Fig. 1b-f). Likely, ATC is not produced in large excess 43 , 26 and thus, plants appear specially sensitive to ATC levels, which exert highest control over 27 pyrimidine de novo synthesis. The production of ATC is under transcriptional regulation in 28 response to tissue pyrimidine availability 42,43 and to growth signals mediated by the TOR pathway 29 45 . However, transcription and translocation of newly synthesized ATC into the chloroplast are slow and energetically costly processes that do not correct for rapid fluctuations needed to 1 maintain nucleotide homeostasis. For this, allosteric regulation by UMP is the major mechanism 2 controlling ATC activity in plants 7 . However, until know we lacked detailed information of how this 3 feedback loop happened. The unprecedented structure of an ATC from plants reveals the mechanism of UMP 7 inhibition. Rather than occupying an allosteric pocket, UMP binds and blocks the active site ( Fig.   8 2), directly competing with CP, the substrate binding in first place 18,46 . The capacity to bind UMP  Table 2). Since the sequence of the CP-loop appears invariant 13 in all plant ATCs known up to date ( Supplementary Fig. 1), we propose that the UMP-inhibition 14 mechanism described here for Arabidopsis ATC must be conserved across the plant kingdom.

15
Thus, unlike other ATCs that rely on complex associations with regulatory proteins, the current

19
The structure also explains the unsolved problem of why plant ATCs are selectively 20 inhibited by UMP and not by any other nucleotide 8,35 . UMP blocks the subunit in a wide-open 21 conformation ( Supplementary Fig. 3), where the N-and C-domains cannot move further apart to accommodate a di-or tri-phosphorylated nucleotide. Also, the pocket for the nitrogenous base is 23 too small for the double ring of a purine and highly selective for uracil, since the methyl group of 24 thymine would clash with the 349 PLP 351 loop, whereas the cytidine amino group would distort the 25 interaction with R310 (Fig. 2e). Finally, one would expect the binding of deoxy-UMP to be weak 26 based on the interaction between the ribose OH groups and the side chain of R248, which mimic 27 the recognition of the Asp a-COOH group (Fig. 2e).

28
atATC has a surprising affinity for UMP, ~400-fold higher than for CP (Table 2), which is  Table 2). Since PALA is a transition-state analog, this finding 26 indicates that catalysis in plant ATCs might occur only in one subunit of the trimer at a time. This 27 unexpected mechanism relies on the projection of F161 towards the threefold axis, which 28 prevents the CP-loops from reaching simultaneously the active conformation (Fig. 4b,d). Thus,

29
F161 plays a dual role both stabilizing the UMP inhibited conformation and synchronizing the firing of the subunits in the trimer. Indeed, mutation F161A does not only turn the enzyme 1 insensitive to UMP but also allows the binding of three PALA molecules per trimer (Fig. 5e), 2 although only one does it with high affinity ( Table 2), suggesting that other elements might 3 contribute to the communication between subunits. On the other hand, it is tempting to speculate 4 that the activation of the subunits could follow a specific order. Certainly, the atATC structure with 5 PALA and CP provides a suggestive snapshot of each subunit at a different stage of the reaction 6 ( Fig. 4c), and further studies should explore this possibility.

7
The alternation in the firing of the active sites might not be exclusive of plant ATCs.

8
Indeed, human and E. coli ATCs exhibit much lower affinities for the binding of PALA to the third 9 active site 28,49 , and this led us to suggest that only two subunits per trimer were catalytically 10 competent at a time 28 . The current results further stress the idea that ATCs in general might work 11 more efficiently if not all subunits react simultaneously, avoiding intersubunit contacts that could 12 slow down the conformational movements needed during catalysis 28,50 (Fig. 3c). Indeed, the 13 obstruction between subunits would explain the decreased activity at high substrate 14 concentrations, a well-characterized but poorly understood phenomenon in ATCs 50 . Interestingly, 15 atATC does not exhibit substrate inhibition (Fig. 4d), in agreement with the existence of a 16 mechanism to prevent the closure of more than one subunit at a time. In turn, F161A bypasses 17 this mechanism and shows substrate inhibition (Fig. 5b,c), although less acute than other ATCs 18 46,50 , which might explain why the activity is not higher than WT despite having three active sites 19 that could fire simultaneously.

22
The de novo pyrimidine synthesis pathway is a promising target for biomedical and

19
ATC (pyrB; At3g20330) knock down mutants were generated using an established protocol for 20 gene silencing by artificial microRNA (amiRNA) 70 . An amiRNA targeting pyrB was designed using 21 an online tool (http://wmd3.weigelworld.org). The sequence TAATGACAGGTATATCGGCAG was used for generation of primers, and Gateway™ compatible sequences attP1 and attP2 were 23 added to primers to engineer the amiRNA fragment (Supplementary Table S1). Subsequently,     To remove UMP from the purified protein, the sample was diluted to 1 mg·ml -1 and supplemented 7 with 50 mM CP and 100 mM Asp, and filtered at room temperature through three consecutive 8 PD-10 desalting columns (GE Healthcare) equilibrated in GF buffer containing 50 mM CP and 9 100 mM Asp. In between columns, the 3.5 ml eluted sample was concentrated down to 2.5 ml 10 using an Amicon ultracentrifugation device. In the last step, the sample was filtered through a PD-