Structure and function of the pseudouridine 5’-monophosphate glycosylase PUMY from Arabidopsis thaliana

ABSTRACT

Pseudouridine is a noncanonical C-nucleoside containing a C-C glycosidic linkage between uracil and ribose. In the two-step degradation of pseudouridine, pseudouridine 5’-monophosphate glycosylase (PUMY) is responsible for the second step and catalyses the cleavage of the C-C glycosidic bond in pseudouridine 5’-monophosphate (ΨMP) into uridine and ribose 5’-phosphate, which are recycled via other metabolic pathways. Structural features of Escherichia coli PUMY have been reported, but the details of the substrate specificity of ΨMP were unknown. Here, we present three crystal structures of Arabidopsis thaliana PUMY in different ligation states and a kinetic analysis of ΨMP degradation. The results indicate that Thr149 and Asn308, which are conserved in the PUMY family, are structural determinants for recognizing the nucleobase of ΨMP. The distinct binding modes of ΨMP and ribose 5’-phosphate also suggest that the nucleobase, rather than the phosphate group, of ΨMP dictates the substrate-binding mode. An open-to-close transition of the active site is essential for catalysis, which is mediated by two α-helices, α11 and α12, near the active site. Mutational analysis validates the proposed roles of the active site residues in catalysis. Our structural and functional analyses provide further insight into the enzymatic features of PUMY towards ΨMP.

KEYWORDS:

• Crystal structure
• C-nucleoside
• metalloenzyme
• conformational changes
• substrate fidelity

Introduction

Pseudouridine is one of the most prevalent RNA modifications [1,2] and is a noncanonical C-nucleoside containing a C-C glycosidic linkage between the C5 atom of the nucleobase uracil and the C1’ atom of ribose (Figure 1). In transfer RNAs (tRNAs), as well as noncoding and coding RNAs, pseudouridine has a role in tRNA stability maintenance and messenger RNA translational regulation, together with diverse biological effects [1–3]. Pseudouridine degradation was first reported in Escherichia coli [4,5] and more recently in Arabidopsis thaliana [6]. In A. thaliana, metabolic linkage of pseudouridine degradation and its products to other pathways were proposed (see Fig. 14 in [6]). Pseudouridine produced from RNA breakdown in the vacuole is exported to the cytosol and then into the peroxisome, in which pseudouridine degradation occurs via a two-step reaction (Figure 1). In the first reaction, pseudouridine is converted into pseudouridine 5’-monophosphate (ΨMP) by pseudouridine kinase (PUKI; YeiC in E. coli) and subsequently ΨMP is cleaved into uracil and ribose 5’-phosphate (R5P) by pseudouridine 5’-monophosphate glycosylase (PUMY; YeiN in E. coli) [5,6]. The products, uracil and R5P, are recycled via other metabolic pathways. Uracil enters a salvage reaction or pyrimidine ring catabolism, and R5P a phosphoribosyl transfer reaction or a pentose phosphate pathway. A bioinformatics search identified the genes for PUKI and PUMY in many bacteria and some eukaryotes [7], but not in mammals, consistent with the observation that pseudouridine is not metabolized in mammals. In A. thaliana, malfunction of pseudouridine catabolism caused adverse physiological effects on plant development and growth [6].

Figure 1. Two-step degradation reaction of pseudouridine. PUKI and PUMY are responsible for pseudouridine catabolism. Atomic numbering of the nucleobase is indicated.

Structural and biochemical studies of PUKI from A. thaliana [8] and E. coli [9] have shown that a threonine or serine residue contributes to the active site of the adjacent subunit in a dimeric enzyme and mediates interactions with the pseudouridine-specific N1 atom. These interactions trigger the conformational changes required for catalysis, providing mechanistic insight into the substrate specificity of PUKI, which has greatly enhanced catalytic efficiency towards pseudouridine over a structurally homologous uridine, despite the similar K_m values for the two nucleosides. Unlike PUKI, which lacks sequence conservation from E. coli to eukaryotes [6,7,9], PUMY exhibited high sequence homology among many bacteria and some eukaryotes (Supplementary Fig. S1) [7], with ~50% sequence identity between E. coli and A. thaliana. PUMY is therefore a distinct enzyme family and catalyses a reversible reaction for the cleavage and formation of the C-C glycosidic bond in ΨMP to and from uridine and R5P, respectively (Figure 1) [7]. In vitro, PUMY strongly favours the reaction for ΨMP synthesis, not ΨMP degradation, with equilibrium constants for the degradation reaction of 8.9 × 10⁻⁵ M [10] and 2.3 × 10⁻⁴ M [5].

The structure of E. coli PUMY (EcPUMY; YeiN) [11] showed the features of the enzyme and a novel catalytic mechanism for the C-C bond cleavage between the nucleobase (hereafter referred to as uracil-Ψ) and R5P. Here, we present three crystal structures of A. thaliana PUMY (AtPUMY), each in a different ligation state. We have also conducted a kinetic analysis of the putative active site residues for the ΨMP degradation reaction. Our findings provide structural and functional insights into the active site residues of PUMY.

Results and discussion

Structure of AtPUMY–Mn²⁺ holoenzyme

For the crystallographic analysis, a C-terminal His-tagged AtPUMY was crystallized by removing the N-terminal 20 residues (Glu21–Met330; Supplementary Fig. S1). In the crystal structure of the AtPUMY–Mn holoenzyme, the asymmetric unit in the space group P2₁3 contains a monomeric AtPUMY (Table 1). Crystallographic symmetry operation of those asymmetric units generated a trimeric structure of AtPUMY, consistent with the results of size-exclusion chromatography (Supplementary Fig. S2A).

Table 1. Data collection and refinement statistics.

Data set	AtPUMY (Mn^2+)	AtPUMY(K185A) (Mn^2+, citrate)	AtPUMY(K185A) (Mn^2+, ΨMP, R5P)
PDB ID	8K05	8K07	8K06
Data collection
Crystal Wavelength (Å)	Native 0.97933	Native 0.97934	Native 0.97932
Resolution (Å)	50.0–1.45 (1.50–1.45)^a	50.0–1.12 (1.16–1.12)	50.0–1.86 (1.93–1.86)
Unique reflections	104,528	316,141	81,648
Multiplicity	21.9 (21.9)	1.9 (1.9)	3.5 (3.5)
Completeness (%)	99.9 (100.0)	92.4 (82.1)	99.0 (99.9)
Mean I/sigma(I)	37.2 (1.3)	16.9 (1.0)	13.0 (2.6)
Wilson B-factors (Å^2)	20.7	12.6	29.0
R-merge^b CC1/2^c	0.07 (2.74) 0.99 (0.46)	0.05 (0.47) 0.99 (0.59)	0.14 (1.11) 0.97 (0.49)
Space group	P213	P1	P21
Unit cell a, b, c (Å)	121.1, 121.1, 121.1	57.3, 68.5, 73.1	72.6, 105.8, 73.6
α, β, γ (º)	90, 90, 90	117.9, 94.3, 109.4	90, 118.9, 90
Refinement
R-work^d (%)	20.6	20.5	18.3
R-free^e (%)	21.7	21.9	20.8
No. of atoms
Macromolecules	2,260	6,623	6,740
Ligands	11	42	17
Water	300	809	646
RMS(bonds) (Å)	0.006	0.006	0.009
RMS(angles) (º)	1.059	1.153	1.197
Ramachandran
Favored (%)	99.01	98.76	98.45
Outliers (%)	0.00	0.34	0.11
Average B-factor (Å^2)	30.7	19.7	33.5
Macromolecules	29.6	18.9	32.7
Ligands	44.5	16.1	45.4
Water	38.9	26.6	42.0

^aNumbers in parentheses refer to data in the highest resolution shell.

^bR_merge =Σ|I_h - <I_h>|/Σ I_h , where I_h is the observed intensity and <I_h> is the average intensity.

^cThe CC_1/2 is the Pearson correlation coefficient (CC) calculated from each subset containing a random half of the measurements of unique reflection.

^dR_work = Σ ||F_obs|-|F_cal||/Σ|F_obs|.

^eR_free is the same as R_obs for a selected subset of the reflections that was not included in prior refinement calculations.

The 1.45 Å resolution crystal structure of the AtPUMY–Mn holoenzyme displayed structural features from Ser28 to Met330, with three additional Leu, Glu, and His residues at the C-terminus, which were introduced for gene cloning. Monomeric AtPUMY adopted an α/β-fold with a twisted central β-sheet (Figure 2A and Supplementary Fig. S1), which exhibited convex and concave sides, with two β-strands (i.e., β5 and β6) and 13 α-helices flanking the β-sheet. Among the flanking elements, two antiparallel helices at the C-terminal region, α11 and α12, protruded from the β-sheet as much as ~34 Å and ~21 Å, respectively, and were highly bent by ~90° towards the concave side of the central β-sheet. The flanking structural elements were topologically dissected into two regions, one with the N-terminal region of Ser28–Gln145 (i.e., α1–α6) and the other with the C-terminal region of Val147–Met330. Between the two regions was a 44 Å-long C-terminal α13. On the concave side of the β-sheet, there is a vacancy between the two flanking regions. The vacancy, which had a wide opening, was surrounded by three flanking clusters, comprising N-domain antiparallel α2 and α3; a protruding loop (magenta in Figure 2A) connecting β7, α7, and α8; and C-domain α11, α12, and α13 (Figure 2A). C-domain α11 and α12 likely serve as a lid for the vacancy. A metal-binding site lies in the ~10 Å inside of the vacancy and along the loop after α7 (Figure 2B). The putative Mn²⁺ ion is of square bipyramidal geometry, with Asp164 and a water molecule across Asp164 in the axial ligand, and four water molecules at the equatorial ligands. Within 2.4 Å of the axial water molecule, one molecule of sulphate ion, which is from the crystallization solution, was identified farther inside the active site, without forming any possible direct coordination to the Mn²⁺ ion. The binding of sulphate was stabilized by interactions at distance less than 3.0 Å with Arg116, Ser166, and Ser167 and with the axial and equatorial water molecules in the metal-binding site. Based on these structural features, the vacancy could serve as an active site of AtPUMY, but the opening remained wide in the monomer (Figure 2B and Supplementary Fig. S3A).

Figure 2. Structural features of the AtPUMY holoenzyme.

A. Monomeric structure of AtPUMY. The concave side of AtPUMY is shown, with the central β-sheet in blue. The figure orientations of A, B, and D are almost identical and corresponded to the left view of subunit A in Figure 2C (i.e., the front view of a trimer). Six α-helices (α1–α4, α10, and α13) were on the convex side, and α-helices α5–α9 and two β-strands were on the concave side. The putative Mn²⁺ ion and one sulphate ion are indicated by a black sphere and a space-filling model, respectively. The loop connecting β7, α7, and α8 is shown in magenta.

B. Zoom view of the metal-binding site in a monomer with an orientation in (A). Asp164 and a water molecule (green circle) across Asp164 are the axial ligands, and four water molecules are equatorial ligands. The dashed lines indicate possible coordination bonds to the metal ion and show a square bipyramidal geometry. Residues possibly stabilizing the binding of sulphate ion are indicated.

C. Front view of an AtPUMY trimer. Each subunit in the surface representation is coloured differently, and colour codes are maintained unless otherwise specified. The three-fold symmetry axis runs through the centre.

D. Active site of AtPUMY in a trimer. An α9* from the adjacent subunit (i.e., subunit C) participates in the active site of subunit A, which generates a pocket-shaped active site in the trimer. The black sphere represents the putative Mn²⁺ ion. The dashed line represents the Cα-interatomic distance between Arg116 at α5 and Asn286 at α12, which indicates the movement of α12 in response to the binding of ligands (see Figure 3D).

Figure 3. Ligand-binding modes and conformational changes in AtPUMY.

A. Citrate-binding mode in the AtPUMY(K185A)–Mn²⁺–citrate complex. This orientation is almost identical to that in Figure 2B. Citrate directly interacts with the Mn²⁺ ion, and residues in the vicinity of citrate, including the side chain of Thr289, are indicated. For clarity, possible hydrogen bonds and coordination bonds at the metal-binding site are indicated by solid lines, and α12 is absent from this presentation. An identical presentation in the presence of α12 is shown in Supplementary Fig. S4A.

B. ΨMP in the active site of the AtPUMY(K185A)–Mn²⁺–ΨMP/R5P complex. Residues, including the side chain of Thr289, within ~4 Å of ΨMP are indicated. Details are similar to those of (A). Note the water molecule between Thr149 and uracil-Ψ. An identical presentation in the presence of α12 is shown in Supplementary Fig. S4B.

C. R5P in the active site of AtPUMY(K185A)–Mn²⁺–ΨMP/R5P complex. R5P has a binding mode different from that of ΨMP. An equatorial water molecule was not identified in this structure. The presentation in the presence of α12 is also shown in Supplementary Fig. S4C.

D. Conformational changes in the active site. The lid elements of α11 and α12 underwent open-to-closed conformational changes in response to the binding of ligands, including citrate, R5P, and ΨMP. An open conformation (i.e., structure in grey) corresponds to the structure of the AtPUMY holoenzyme in Figure 2A; the three subunits in the AtPUMY(K185A) – Mn²⁺–ΨMP complex represent the closed conformation (purple).

In trimeric AtPUMY, intersubunit interactions transformed the putative, open active site in a monomer into a pocket-shaped site. Three subunits were arranged with three-fold symmetry, and each was oriented with the convex side outwards (Figure 2C). In the trimer, α9* (herein, an asterisk represents residues or elements from the nearby subunit) on the concave sides of β-sheet mediated intersubunit interactions with the flanking elements on the concave sides of the nearby subunits (Figure 2D), which provides an additional wall to the vacancy, thus generating a pocket-shaped active site in the trimer (Supplementary Fig. S3B). Protein interfaces, surfaces, and assemblies analysis [12] indicated that trimerization buried a surface area of 7770 Ų.

Structural features of AtPUMY(K185A) in complex with citrate or ΨMP

We employed an AtPUMY(K185A) mutant to identify the ligand-binding modes. Lys185 in AtPUMY is equivalent to Lys166 in EcPUMY (Supplementary Fig. S1); it was responsible for the cleavage of the C-C glycosidic bond in ΨMP (see Comparisons with EcPUMY and recognition of ΨMP) [11]. The 1.12 Å resolution crystal structure of the AtPUMY(K185A)–Mn²⁺–citrate complex contained a trimeric structure in the asymmetric unit of the P1 space group (Table 1). The complex showed binding of one molecule of citrate to each subunit, using a crystallization solution containing 0.1 M sodium citrate. Citrate is directly bound to the metal-binding site, with its binding mode differing from that of sulphate in the holoenzyme (Figure 3A and Supplementary Fig. S4A). The C2 hydroxyl group and C2 carboxylate group of citrate directly coordinated the metal ion by replacing an axial and one of the equatorial water molecules, respectively. The distal C1 and C3 carboxyl groups of citrate interacted with Thr289 in α12 and Ser167 in a loop following α7, respectively, and the C2 hydroxyl group mediated an intersubunit interaction with Glu198* in α9×. The C3 carboxyl group was located farther inside the active site at a position coinciding with that of sulphate in the holoenzyme.

Co-crystallization of AtPUMY(K185A) with ΨMP produced a crystal of the P2₁ space group with a trimer in the asymmetric unit (Table 1) and showed two different ligands at 1.86 Å resolution. Substrate ΨMP was present at two subunits (Figure 3B and Supplementary Fig. S4B) and the reaction product R5P was present in one subunit (Figure 3C and Supplementary Fig. S4C), possibly due to residual glycosylase activity of the K185A mutant or a non-enzymatic reaction towards ΨMP. The binding modes of ΨMP and R5P might represent unproductive ones for catalysis due to mutation K185A. We later noticed that ΨMP position is identical to that of a possible reaction intermediate of EcPUMY [11], but that R5P displayed a binding mode that differed from those of the other ligands (see Comparisons with EcPUMY and recognition of ΨMP). The complexes thus provide structural information that facilitates functional analysis. ΨMP and R5P were found in the active site with an orientation almost parallel to α12 (Supplementary Fig. S4B,C). The phosphate groups of the ligands were located near the metal-binding site, forming interactions with water molecules in the metal-coordinating shell. The ribosyl moiety and nucleobase uracil-Ψ in ΨMP were aligned towards the C-terminal α13, the ceiling element for the active-site pocket. There are several notable features, including the binding mode for ligands and conformational changes. First, there were no direct interactions of the phosphate group of ΨMP and R5P to the metal ion. In contrast to the citrate-binding mode, the metal-coordinating environments in this complex are identical to those in the holoenzyme. The 5’-phosphate group in R5P, which was within ~4 Å of Arg116, Ser166, and Ser167, occupied the site for the sulphate ion in the holoenzyme, and its ribosyl moiety was further stabilized by interacting with Lys112 and Thr289 within ~4 Å (Figure 3C). Second, ΨMP showed a binding mode distinct from that of R5P (Figure 3B). The 5’-phosphate group in a ribosyl moiety of ΨMP occupied a site corresponding to the binding site for the ribosyl C4’ atom of R5P, not the phosphate group, but maintained interactions with water molecules in the metal-coordinating shell. The phosphorus atoms in ΨMP and R5P showed ~3.3 Å displacement (Figure 3B,C). Thus, the uracil-Ψ moiety of ΨMP was located in the ceiling of the active site. The side chains of the following residues were within ~5 Å of each moiety in ΨMP: Lys112, His156, Ser166, Ser167, Glu198*, and Thr289 from the 5’-phosphate group; Glu50, Thr131, and Asp168 from the ribosyl moiety; and Thr52, Ile53, Phe215, Leu293, and Asn308 from uracil-Ψ. In particular, hydrophobic features, except for Asn308 at the ceiling element α13, were notable in the active site around uracil-Ψ, with one or two water molecules between uracil-Ψ and Thr149 (Figure 3B). We later noted that Thr149 likely plays a role in recognizing the uracil-Ψ moiety in ΨMP (see Comparisons with EcPUMY and recognition of ΨMP).

The structures of these ternary complexes exhibited high similarities with that of the holoenzyme, with a root-mean-square deviation of 0.89–1.27 Å for the 267 Cα atoms of AtPUMY(K185A)–Mn²⁺–citrate and 0.83–1.13 Å for the 272 Cα atoms of AtPUMY(K185A)–Mn²⁺–ΨMP and R5P. Despite the high homology in the overall structures, we observed ligand-dependent structural changes of the lid, which consisted of α11 and α12, to the active site pocket. We adopted the Cα-interatomic distance between Arg116 at α5 and Asn286 at α12, which indicated the movement of the lid (Figure 2D and 3D). The Cα-interatomic distance was ~20 Å in the holoenzyme, ~15 Å in the complex with citrate, and ~14 and ~12 Å in the complexes with R5P and ΨMP, respectively, indicating that the lid closed the pocket upon binding of ligands, and the phosphate moiety and a uracil-Ψ base in the ligands contribute to the lid closure. There is a direct interaction of Thr289 at the lid element α12 with the citrate C1 carboxyl group in the complex with citrate (Figure 3A), to a ribosyl moiety of R5P and the 5’-phosphate group of ΨMP in the complexes with R5P and ΨMP (Figure 3B,C), respectively, together with hydrophobic interactions between ΨMP and residues in α12. These interactions possibly stabilize the closed conformation of the active site of AtPUMY.

Kinetic analysis

We conducted a steady-state kinetic assay of PUMY activity with full-length AtPUMY (Figure 4A); the K_m and k_cat values for ΨMP degradation were 2.3 µM and 62.8 min⁻¹. Subsequent assay indicated that AtPUMY requires the Mn²⁺ ion for full activity, with the K_m value of 2.3 µM (Supplementary Fig. S5A,B). Consistent with the C-nucleoside glycosylase activity of EcPUMY [11], AtPUMY was also specific for ΨMP thus exhibiting, essentially, no measurable activity on other purine and pyrimidine N-nucleoside monophosphates (Figure 4B). We next performed activity assays and kinetic measurements to identify the structural determinants of ΨMP degradation. Seventeen residues within ~5 Å of each moiety in ΨMP (Figure 3B) were mutated, and the mutants subjected to functional assays. Specific activity measurements indicated that most residues contributed to activity, with 0–17% of the wild-type enzyme activity (Figure 4C), except for H156A (74%) and S167A (99%). Subsequent kinetic analysis of the mutants also revealed that mutation of most residues significantly reduced catalytic efficiency (i.e., k_cat/K_m). The resulting catalytic efficiencies of mutants showing measurable activity were 0.5 ~ 4.5% relative to the wild-type enzyme, except for T52A (19.8%), S133A (31.8%), H156A (26.6%), and S167A (55.3%) (Table 2); all exhibited 0.5 ~ 27-fold increases in K_m values and 0.7 ~ 38.5-fold reductions in k_cat values. We also found that certain mutants (i.e., E50A, K112A, D164A, S166A, K185A, L293G, and T289A) were inactive even at enzyme concentrations up to 100-fold (i.e., 8 µM) of other mutants. Such inactivities closely reflected the functional roles. Glu50 corresponds to Glu31 of EcPUMY, which may function as a proton donor in catalysis (see Comparisons with EcPUMY and recognition of ΨMP) [11]. Lys112 and Ser166 are phosphate-binding sites of the substrate ΨMP, and Leu293 serves as a binding site for the uracil-Ψ moiety in ΨMP. Asp164 is a metal-coordinating residue (Figure 2B), suggesting that the metal-binding site anchors the phosphate group of the substrate, even in the absence of any direct interactions between them. The role of Thr289 in α12 is to mediate interactions with the phosphate group in ligands and to accompany movement of the lid helix α12 in response to the binding of ligands. The inactive T289A mutant suggests that ligand-dependent movement of the lid region is required for catalysis.

Figure 4. Enzymatic properties of AtPUMY.

A. Steady-state kinetic analysis of wild-type AtPUMY. Using ΨMP enzymatically synthesized as the substrate, the reaction product R5P was quantified using an enzyme-coupled assay. Each measurement was conducted in triplicate. Experimental details are provided in the Materials and methods.

B. AtPUMY activity towards C- and N-nucleosides. Reactions were performed as in (A) with 25 µM nucleoside 5’-monophosphates and a 60 s reaction. n.d., not detected.

C. Specific activities of mutant AtPUMYs towards 25 µM ΨMP. Reaction details are identical to (B), but enzyme concentrations were increased from 80 nM to 400 nM and 8 µM if no activity was detected. Each measurement was conducted in triplicate. n.d., not detected with 8 µM AtPUMY. Mutants were grouped according to their relative locations to each moiety in ΨMP (see Table 2), and the inactive mutants were also grouped independently.

Table 2. Kinetic parameters of AtPUMY mutants.

		Km (μM)^a	kcat (min^−1)	kcat/Km (μM^−1 min^−1)
Wild type AtPUMY		2.30 (0.29)^b	62.8 (2.4)	27.3	100%
Mutants in the vicinity of the uracil-Ψ	T52A	0.964 (0.30)	5.22 (0.27)	5.41	19.8%
	I53A	3.96 (3.2)	1.63 (0.32)	4.11 × 10^−1	1.51%
	T149A	34.9 (12)	7.67 (0.79)	2.20 × 10^−1	0.81%
	T149V	61.1 (22)	8.25 (1.1)	1.35 × 10^−1	0.50%
	F215A	37.4 (16)	18.2 (0.05)	4.86 × 10^−1	1.78%
	L293G	n.d ^c	n.d.	n.d.	n.d.
	N308A	15.9 (5.7)	2.27 (0.29)	1.43 × 10^−1	0.52%
Mutants in the vicinity of the ribosyl moiety	E50A	n.d.	n.d.	n.d.	n.d.
	T131A	14.8 (3.3)	18.2 (0.91)	1.23	4.51%
	S133A	1.98 (1.1)	17.2 (1.9)	8.68	31.8%
	K185A	n.d.	n.d.	n.d.	n.d.
Mutant in the vicinity of the 5’-phosphate ΨMP	K112A	n.d.	n.d.	n.d.	n.d.
	H156A	12.4 (1.8)	90.0 (3.9)	7.26	26.6%
	S166A	n.d.	n.d.	n.d.	n.d.
	S167A	5.62 (0.83)	84.9 (2.8)	15.1	55.3%
	E198A	62.5 (11)	51.1 (2.6)	8.18 × 10^−1	3.00%
	T289A	n.d.	n.d.	n.d.	n.d.
Mutant for the metal coordinating residue	D164A	n.d.	n.d	n.d.	n.d.

^aThe K_m values were measured towards ΨMP.

^bThe numbers in parentheses represent the standard errors.

^cn.d., not detected with 8 µM AtPUMY.

Interpreting the kinetic parameters of the mutants is difficult, given that PUMY-mediated catalysis features multiple steps [11]. The locations of the residues (Figure 3B,C) and the catalytic efficiencies of the mutants (Table 2) suggest possible functional roles of the residues of interest. In the vicinity of the uracil-Ψ moiety in ΨMP, Thr149 and Asn308 may play an essential role in recognizing the uracil-Ψ moiety and catalysis, via hydrogen bonding, and Ile53 and Phe215 create a hydrophobic environment for uracil-Ψ moiety. Relative to the catalytic efficiency of the wild-type enzyme, the figures for the mutants were; T149A (0.8%), T149V (0.49%), N308A (0.52%), I53A (1.50%) and F215A (1.78%). Thr131 and Ser133 likely serve as the binding sites for the ribosyl moiety and His156 and Ser167 as phosphate-binding sites. Glu198 is unique in that it originates from an adjacent subunit, but nonetheless plays an important role as a phosphate-binding site of ΨMP.

Comparisons with EcPUMY and recognition of ΨMP

The structural features reported here are similar to those of EcPUMY [11]. A previous structural analysis of EcPUMY identified two possible reaction intermediates: a ring-opened ribose ΨMP adduct (Supplementary Fig. S4D) and a ring-opened R5P adduct, but ΨMP was identified as a free form in the EcPUMY(K166A) mutant. The binding modes of the two adducts and ΨMP were identical, particularly for the phosphate group. Each adduct was covalently attached to Lys166^EcPUMY (superscript EcPUMY and AtPUMY represent those of EcPUMY and AtPUMY, respectively), providing a structural clue to the ribose ring-opening mechanism for the cleavage of the C-C glycosidic linkage in ΨMP. In the proposed mechanism, a ribose ring-opening reaction involves the proton donor Glu31^EcPUMY and is followed by glycosyl bond cleavage, which is facilitated by the formation of a covalent linkage between ring-opened ribose and Lys166^EcPUMY. During C-C glycosidic bond cleavage, Thr130^EcPUMY, which corresponds to Thr149^AtPUMY, was suggested to stabilize the leaving uracil by forming a hydrogen bond to the uracil-Ψ N1 atom. In a ring-opened ribose ΨMP adduct, a uracil-Ψ repositioned its orientation by rotating ~90° from that of ΨMP and was located farther inside, where the pseudouridine-specific N1 atom is within a hydrogen-bonding distance of ~3.5 Å from Thr130^EcPUMY (Supplementary Fig. S4D) [11]. Further kinetic analysis validates the proposed functional roles of these active site residues for ΨMP synthesis [11].

In this study, we characterized that the binding mode of free ΨMP to AtPUMY was identical to those of EcPUMY including a ring-opened ribose ΨMP adduct (Supplementary Fig. S4D) and a free ΨMP in the K166A mutant, except for the orientation of uracil-Ψ. The pyranosyl R5P of AtPUMY exhibits a binding mode distinct from those of other ligands (Figure 3B,C), indicating that the reaction product R5P readily departs from the active site. Our structure-based mutational analysis of AtPUMY affords further insights into the functional roles of residues around ΨMP. The AtPUMY-mediated catalysis exhibits several features of interest. First, Thr149 and Asn308 near uracil-Ψ are likely major determinants for recognizing the uracil-Ψ of ΨMP. It is known in EcPUMY that Thr130^EcPUMY recognizes the uracil-Ψ N1 atom in a ring-opened ribose ΨMP adduct [11], strongly supporting our kinetic analyses of T149A and T149V mutants (Table 2). Second, the sequence conservation (Supplementary Fig. S1) between EcPUMY [11] and AtPUMY and their structural features both indicate that hydrophobic properties and a spacious active site to accommodate water molecules around uracil-Ψ are general features conserved in the active site of PUMY (Supplementary Fig. S4D). These characteristics allow the nucleobase uracil-Ψ to reorient during catalysis. If there were specific or strong interactions around uracil-Ψ at the beginning of catalysis, uracil-Ψ repositioning could not be achieved. For example, a rotation of ~180° around the NE–CZ bond in arginine is required for catalysis in arginine dihydrolase [13,14], which consecutively liberates two ammonia molecules from the guanidinium group of arginine. In arginine dihydrolase, asparagine is essential for orienting the susceptible guanidinium group of arginine to a catalytic residue. Mutation of asparagine to aspartate eliminated activity, indicating that relatively weak interactions with the substrate allow the proposed rotation of a reaction intermediate. Third, the binding modes of ΨMP and R5P phosphate groups to AtPUMY present additional features (Figure 3B,C). It is unusual in metalloenzymes that the phosphate group of ligands mediates interactions only with the metal-coordinating water molecules, which was commonly identified in both EcPUMY [11] and AtPUMY. Apart from the unusual indirect bindings of phosphate groups to metal ion in AtPUMY, the different binding locations of the ΨMP and R5P phosphates are also unique. Such results suggest that ligand phosphate groups do not serve as major anchors in terms of substrate-binding to PUMY; rather, a nucleobase in the substrate dictates the productive binding mode for catalysis. This explains why the binding site of the reaction product R5P of the AtPUMY (K185A) mutant differed from that of the phosphate group in ΨMP (Figure 3B,C). However, the phosphate group-binding site of a ring-opened R5P adduct in EcPUMY was identical to those of a ring-opened ribose ΨMP adduct and a free ΨMP, given the covalent linkage to Lys166^EcPUMY. Conformational changes in the lid region are an additional feature in catalysis (Figure 3D). Conformational changes, including the repositioning of uracil-Ψ in the substrate and lid movement in the enzyme, are required for catalysis, as suggested by the catalytically impaired T149A and T149V mutants, and the inactive T289A mutant, respectively.

Conclusions

Structural and functional analyses of AtPUMY suggest that a nucleobase uracil-Ψ is required for ΨMP to bind to the active site in a productive mode for catalysis, and a concurrent transition of the active site into a closed state by lid movement is also essential for catalysis. Recognition of the uracil-Ψ N1 atom and the nucleobase release are likely facilitated by Thr149. Therefore, mutation of Thr149 rendered AtPUMY an inefficient enzyme, with <1% of the catalytic efficiency of wild-type AtPUMY. Our findings provide the structural and biochemical bases of AtPUMY catalysis towards ΨMP.

Materials and methods

Cloning and purification

A gene for a full-length AtPUMY (Met1–Met330; GenBank: AT1G50510) was cloned from the cDNA library of A. thaliana. For the structural and biochemical studies, genes for various AtPUMYs were produced by PCR with the respective primers (Supplementary Table s1).

For crystallographic analysis, crystallization of C-terminal His-tagged AtPUMY was achieved by truncating the N-terminal 20 residues. The resulting variant of C-terminal His-tagged AtPUMY (Glu21–Met330) was expressed and purified as follows: The PCR product of AtPUMY was cloned into the pET-41b expression vector (Merck, Kenilworth, NJ, USA) containing a C-terminal (His)₅-tag. E. coli Rosetta2 (DE3) cells (Merck) transformed with constructs for AtPUMY were cultured at 37°C in Luria – Bertani medium until the absorbance at 600 nm reached 0.5–0.7. Enzyme expression was induced by adding 0.5 mM IPTG, followed by incubation at 20°C for ~15 h. Cells were collected by centrifugation and sonicated in buffer A (50 mM Tris at pH 8.0, 300 mM NaCl) and 1 mM MnCl₂, and the supernatant was obtained by centrifugation at 30,000 × g for 30 min at 4°C. The C-terminal His-tagged enzyme was purified using a HisTrap HP column (GE Healthcare, Chicago, IL, USA) with buffer A and eluted using buffer B (buffer A plus 500 mM imidazole) and subsequently subjected to size-exclusion chromatography using a HiLoad 16/60 Superdex 200 pg column (GE Healthcare) with buffer A.

For biochemical analysis, the wild type and various mutants of full-length AtPUMY, with a C-terminal His-tag, were constructed. The AtPUMYs were purified using a HisTrap HP column as described above followed by further purification with a HiPrep 26/10 Desalting column (GE Healthcare) with buffer A. All the enzymes were stored at −20°C.

Crystallization and structure determination

Purified AtPUMY and AtPUMY(K185A) mutants in 50 mM Tris at pH 8.0 and 300 mM NaCl were concentrated to ~16 and ~44 mg mL⁻¹ and subjected to crystallization using the sitting drop vapour diffusion method at 22°C. Crystallization of AtPUMY in the presence of Mn²⁺ was achieved, first by crystallizing an apo-AtPUMY, with a solution containing 0.1 M cacodylate (pH 6.5) and 1.4 M ammonium sulphate, and then by soaking a crystal of apo-AtPUMY for 1 min with 8 mM MnCl₂. A crystal of AtPUMY(K185A) complexed with Mn²⁺ and citrate was obtained by crystallization of AtPUMY(K185A) in the presence of 2 mM Mn²⁺, 2 mM R5P, and 2 mM uracil, with 0.1 M sodium citrate pH 6.5 and 16% PEG 10,000. Later, we noticed under these crystallization conditions that the structure of AtPUMY(K185A) contained Mn²⁺ and citrate from the crystallization solution, not a reaction product. Crystallization of the AtPUMY(K185A) mutant in a complex with ΨMP was carried out in the presence of 100 µM ΨMP and 2 mM MnCl₂ using a solution of 0.2 M lithium sulphate, 0.1 M MES pH 6 and 20% (w/v) PEG 4000. The substrate ΨMP was enzymatically synthesized (see Enzyme-dependent synthesis of ΨMP). In all cases, 20% ethylene glycol was used as a cryoprotectant for data collection at 100 K. Diffraction data were collected with a 0.5° oscillation angle at beamline 7A and 11C of the Pohang Accelerator Laboratory (Pohang, Republic of Korea). Program HKL2000 suite was used for diffraction data processing [15] and a CC_1/2 statistical value of approximately 0.4–0.5 was used as the high-resolution cut-off [16,17].

Three structures had a high resolution of 1.12–1.86 Å. Crystals of the AtPUMY–Mn²⁺ complex belonged to the space group P2₁3, with a monomer in the asymmetric unit, and two complexes of AtPUMY(K185A)–Mn²⁺–citrate and AtPUMY(K185A)–Mn²⁺–ΨMP/R5P in the space groups P1 and P2₁, respectively, with a trimer in the asymmetric unit. Structural determination of AtPUMY was carried out by molecular replacement using PHENIX AutoMR [18], with EcPUMY (PDB id, 4GIK) [11] as the search model. The refined model of the AtPUMY holoenzyme was used as a search model for molecular replacement of other complexes. The models were manually rebuilt by COOT [19,20] and refined with PHENIX. Details on data collection and refinement statistics are shown in Table 1.

Enzyme-dependent synthesis of ΨMP

ΨMP was synthesized via a PUKI-dependent reaction, with pseudouridine as the substrate (Figure 1). PUKI from E. coli was prepared as described previously [9]. The ΨMP synthesis reaction was carried out at 25°C for 20 min in 50 mM Tris-HCl (pH 7.5), 10 mM pseudouridine, and 1 µM PUKI, with 20 mM MgCl₂, 50 mM KCl, and 15 mM ATP. After 20 min, PUKI was filtered out using an Amicon centrifugal filter (EMD Millipore, Burlington, MA, USA) with a molecular weight cut-off of 30 kDa (Merck). The formation of ΨMP was assessed by high-performance liquid chromatography (HPLC) as described below (Supplementary Fig. S2B). The ΨMP-containing reaction mixture was stored as aliquots at −80°C.

The Thermo Scientific™ UltiMate™ 3000 HPLC system (Thermo Fisher Scientific, Waltham, MA, USA) was used to characterize ΨMP from PUKI-dependent reaction mixtures. For HPLC, the enzymatic reaction was carried out for the PUKI-dependent reaction described above. After 20 min, instead of removing PUKI with a filter, we quenched ΨMP synthesis by adding an equal volume of methanol. After thorough mixing, the reactants and products were collected for HPLC. The HPLC analysis was performed as described previously [10]. Analytes were separated using a reverse-phase ion-pairing INNO C18 column (5 µm, 4.6 × 150 mm; YoungJin Biochrom, Gyeonggi-do, Korea), at an oven temperature of 30°C. The flow rate was 1.5 mL min⁻¹ and the injection volume was 2 μL. Isocratic elution for 10 min was performed with a mobile phase containing 50 mM phosphate buffer (pH 5.9), 40 mM tetrabutylammonium bromide and 12.5% acetonitrile. The analytes were detected using a diode array detector at 260 nm.

Enzyme-coupled assay for AtPUMY

R5P, a reaction product of AtPUMY, was continuously monitored using an enzyme-coupled assay [5]. In the assay, ribose-5’-phosphate isomerase (RpiA) and an NADPH-dependent ribulose 5’-phosphate reductase (Acs1) catalysed two consecutive reactions – R5P into ribulose 5’-phosphate, and further into ribitol 5’-phosphate – with concurrent changes of absorbance at 340 nm. The steady-state kinetic assay was carried out at 30°C using a UV–Vis spectrophotometer (Jasco, Tokyo, Japan).

The pre-reaction mixture consisted of 50 mM HEPES (pH 7.5), 2 mM MnCl₂, and a given concentration of ΨMP, with coupling enzymes of 5 µM C-terminal His-tagged RpiA and 4 µM N-terminal His-tagged Acs1, and 180 μM NADPH. For maximal Acs1 activity, 2 mM dithiothreitol and 50 μM cytidine triphosphate were added to the pre-reaction mixture. After incubation at 30°C for 120 s, an enzyme-dependent reaction was triggered by adding 80 nM AtPUMY and decreases in absorbance at 340 nm were recorded. We verified in preliminary experiments that the concentrations of the coupling enzymes RpiA and Acs1 and the divalent cation Mn²⁺ were saturating in the assay. The initial velocity was between 15 and 30 s and was converted to the level of R5P using a standard curve. SigmaPlot (Systat Software, San Jose, CA, USA) was used to calculate the values of K_m and V_max.

Author contributions

SR supervised the project; SHK performed crystallographic experiments of holoenzyme and the complex with citrate, and JL determined the crystal structure of the complex with ΨMP/R5P; JL carried out biochemical and kinetic assays; SR and JL analysed the data and wrote the manuscript with contributions from SHK.

Acknowledgments

We thank Minsoo Kim for her comments on a PUKI-dependent ΨMP synthesis and Dr. Ah-Reum Lee for providing us with the cDNA library of A. thaliana.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The atomic coordinates and structural factors have been deposited in the Protein Data Bank (http://www.rcsb.org) under ID code 8K05, 8K06, and 8K07.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15476286.2023.2293340

Additional information

Funding

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) [2021R1A2C2092118 and RS-2023-00207820].