The Structure of Mn^II–Bound Rubisco from Spinacia oleracea

Abstract

The rate of photosynthesis and, thus, CO₂ fixation, is limited by the rate of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco). Not only does Rubisco have a relatively low catalytic rate, but it also is promiscuous regarding the metal identity in the active site of the large subunit. In Nature, Rubisco binds either Mg(II) or Mn(II), depending on the chloroplastic ratio of these metal ions; most studies performed with Rubisco have focused on Mg-bound Rubisco. Herein, we report the first crystal structure of a Mn-bound Rubisco, and we compare its structural properties to those of its Mg-bound analogues.

Graphical Abstract

The 2.5 Å crystal structure of Mn2+-substituted spinach RuBisCO with bound transition state analogue is reported. Structural comparisons are drawn to the Mg2+ analogue, and density functional theory calculations are carried out to attempt to rationalize differences in selectivity for ribulose-1,5-bisphosphate carboxylation vs oxygenation.

Introduction

Rubisco is a metalloprotein that catalyzes the carboxylation of ribulose-1,5-bisphosphate (RuBP) to form 2 equivalents of 3-phosphoglycerate (3-PG), which are used for saccharide synthesis via gluconeogenesis. Thus, Rubisco is the major biological gatekeeper of CO₂ sequestration and fixation¹. Rubisco participates in photorespiration as well, where RuBP is oxygenated to form one equivalent of 3-PG and one equivalent of phosphoglycolate. However, photorespiration is widely regarded as a futile cycle, as the recycling of phosphoglycolate requires the expenditure of 3.5 ATP molecules and two NADPH molecules to recycle phosphoglycolate back into RuBP^2-4.

Structures obtained to date spanning several Rubisco species reveal that its active site is comprised of a distorted octahedral inner coordination sphere about a divalent metallocofactor⁵. In the absence of substrate, this coordination sphere consists of three water molecules, an Asp, a Glu, and a carbamylated Lys⁶. In Spinacea oleracea (spinach) Rubisco, the coordinating amino acids are Asp203, Glu204, and Lys201. When the native substrate, RuBP, is bound, the two equatorial water molecules are replaced by the substrate’s C2 and C3 position oxygen atoms. This leaves a single axial site, bound by water⁷, available to stabilize one of the two carboxyl oxygen atoms in the transition state resulting from the binding of CO₂ to the C2 position of the enol form of bound RuBP (Scheme 1). To glean information about the reactivity of Rubisco, the inhibitor and transition state analog, 2-carboxyarabinitol 1,5-bisphosphate (CABP), is often used to mimic the transition state of Rubisco during the carboxylation reaction, wherein the analogous atoms to RuBP on CABP bind to the same positions irreversibly⁸, providing a rigid scaffold to study the structure of the active enzyme (Figure 1).

Scheme 1.

Tautomerization of RuBP and structure of CABP.

Figure 1.

Structure of the spinach Rubisco active site, highlighting amino acid interactions and binding of CABP (red).

Rubisco is commonly described as a Mg-binding metalloenzyme, as this is widely considered the biologically relevant form. While the Mg-bound form is the most common form in nature, Rubisco frequently binds Mn(II) as an alternative to Mg(II). Either metal ion in the active site is labile, so the relative ratio of each metal ion in Rubisco in vivo is principally affected by the chloroplastic ratio of Mg(II) to Mn(II)⁹. While these are the two canonically biologically relevant divalent metallocofactors, Rubisco is also capable of binding a variety of different divalent metal ions which also activate the enzyme for carboxylation/oxygenation reactivity with RuBP¹⁰.

Before the binding of any metals in the active site, Rubisco must be activated via carbamylation of a specific lysine residue¹¹ (Lys201), which is facilitated by the metal ion. It is unknown if this mechanism of activation is different for Mn, but we consider the prospect of differing mechanisms of activation unlikely given that Rubisco can bind many different divalent metals in its active site and exhibit activity toward CO₂ and/or O₂. When Mn(II) is bound in the active site rather than Mg(II), the substrate specificity changes dramatically. The CO₂ affinity is decreased roughly 4-fold and the O₂ affinity is unaffected^9,12, becoming essentially nonspecific toward either substrate. It has been hypothesized that the poor substrate specificity of Rubisco is largely due to the interplay between stabilizing the carboxy transition state to select for carboxylation over oxygenation but not stabilizing it so much such that the products are not released¹³. This hypothesis, however, does not address the difference in reactivity between Mg- and Mn-bound Rubisco. Probing Mn-bound Rubisco could offer mechanistic and structural insight into how metal identity dictates substrate selectivity.

There are 3 known forms of Rubisco: I, II, and III¹⁴, all of which have high resolution representative structures in the Protein Data Bank (PDB). Type I Rubisco is the most common and is found in bacteria and eukaryotic chloroplasts, including all higher plants. Form I is a hexadecamer composed of 8 large (50-55 kDa) and 8 small (12-18 kDa) subunits, where the active sites are contained in an alpha/beta barrel-like motif in the large subunit, yielding a total of 8 active sites per enzyme. We now report the structure of a Mn-bound form I Rubisco from Spinacia oleracea.

Materials and Methods

General Considerations.

Milli-Q water (18.2 MΩ; Millipore) was used in the preparation of all buffers and solutions. Activity assay data were measured using a Clark-style oxygen electrode (Oxygraph+ system, Hansatech). All chemicals were purchased from VWR International. All reactions were performed in technical triplicate using air-saturated buffer solutions.

Synthesis.

Reaction intermediate and transition state analog, CABP, was synthesized using previously reported procedures⁸ and the product was verified using both NMR and activity assays. The crude reaction product, containing approximately equal parts of both diastereomers of the product, as determined by NMR (Figure S1), was used in all reaction conditions utilizing CABP. It has been previously reported that only the desired product, CABP, is strongly binding and inhibitory^15,16, so we deemed further purification unnecessary.

Protein Purification.

Rubisco from Spinacea oleracea was obtained using commercially available fresh spinach from Wegmans in Ithaca, NY. This spinach was flash frozen in liquid nitrogen and then added to a blender to pulverize the frozen spinach. The powdered spinach was then suspended in a solution of 50 mM tris (pH = 7.4), 20 mM MgSO₄, 20 mM NaHCO₃, 50 mM β-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 10% (v/v) glycerol in a ratio of 1 kg spinach/1 L buffer at 4 °C for 2 h. This extract was then filtered through 8 layers of cheesecloth 3 times to remove large insoluble leaf debris. The smaller leaf debris was removed by centrifugation at 20,000 g. The resulting supernatant was subjected to an ammonium sulfate precipitation where the 30-50% saturation ammonium sulfate cut at 4 °C was kept. This pellet was resuspended in and then further dialyzed into Q-column buffer A (50 mM tris (pH = 7.6), 20 mM MgCl₂, and 20 mM NaHCO₃). This solution was then applied to a Q-Sepharose anion-exchange column and eluted via a gradient over 10 column volumes with buffer B being the same as buffer A brought to 500 mM NaCl. Spinach Rubisco, which exhibited a pale-yellow color, eluted at a conductivity of 27 mS/cm. This protein was homogeneously pure as determined by SDS-PAGE (Figure S2).

Purified spinach Rubisco was then dialyzed into 50 mM tris (pH = 7.6), and 150 mM NaCl. Due to the addition of excess exogenous Mg(II) throughout purification, the protein was assumed to be mostly the Mg-bound form. This solution was treated with 5 mM ethylenediamminetetraacetic acid (EDTA) to remove the Mg(II) from the enzyme to generate apo-Rubisco and applied to a PD-10 desalting column to remove Mg(II)-bound EDTA as well as excess EDTA. This desalted solution was then tested for activity using an Oxygraph+ oxygen electrode, revealing the abolition of oxygenase activity with the native substrate, RuBP. The apo-Rubisco solution was then treated with 10 mM MnCl₂ and 10 mM NaHCO₃ to reconstitute the apo-Rubisco with Mn(II). Excess Mn(II) and NaHCO₃ were then removed via desalting column as described above.

During the metal-stripping and then reconstitution of the natively purified Rubisco, activity assays were performed after each step to ensure expected activity or the lack thereof. Oxygen electrode activity assays were performed using 50 mM tris (pH = 7.6), 150 mM NaCl, 50 μM Rubisco, and 10 mM RuBP in air-saturated water. CABP-bound Rubisco was generated by treating Rubisco with at least a 3-fold excess of CABP-relative to active site concentration-for 30 minutes at ambient temperature, and then desalted to remove excess CABP using a PD-10 desalting column.

Electron Paramagnetic Resonance (EPR) Spectroscopy.

X-Band (9.40 GHz) EPR spectra were recorded on a sample containing 300 μM Mn(II)- and CABP-bound Rubisco in 50 mM tris (pH = 7.6), 150 mM NaCl with 30% (v/v) glycerol. The measurements were taken using a Bruker Elexsys-II spectrometer equipped with a liquid He cryostat maintained at 100.0 K.

Protein Crystallography.

Mn(II)- and CABP-bound Rubisco was buffer exchanged into 50 mM potassium phosphate (pH = 7.0) and crystals were grown using the sitting drop method using a 1:1 (v/v) mixture of 10 mg/ml of the aforementioned Rubisco solution and precipitant solution containing 200 mM lithium acetate (pH = 5.3) and 22% (w/v) PEG 3350. Crystals formed under these conditions after 8 days at 25 °C and these were looped, dipped in a cryoprotectant solution consisting of the precipitation buffer containing 30% glycerol, and flash frozen in liquid N₂.

Crystallographic data were collected at the Advanced Photon Source at Argonne (beamline NE-CAT 24-ID-C) to a high-resolution limit of 2.25 Å, but based on merging statistics, the data were truncated to a resolution of 2.5 Å. A structural model was obtained via molecular replacement as implemented in PHENIX¹⁷ using 8RUC as a starting model. Model building and chain-walking real-space refinements were carried out in COOT¹⁸, with further iterative refinements carried out using Phenix.

Computational Methods.

Density functional theory (DFT) calculations were performed with version 4.2 of the ORCA software package¹⁹. Truncated active site models of Mg- and Mn-bound Rubisco were generated from crystallographic coordinates. These models included all inner-sphere ligands to the divalent metal, and their coordinates are included as Supporting Information. Positions of hydrogen atoms were optimized via hybrid DFT calculations. Both optimizations and single-point calculations used the B3LYP hybrid density functional^20,21. The RIJCOSX²² algorithm was used to speed the calculation of Hartree–Fock exchange. The scalar relativistically recontracted ZORA-def2-TZVP(-f) basis set was used for all atoms²³. Grid integration accuracy was set to ORCA GRID7 for the metal atoms and GRID4 for all other atoms. Calculations included the zeroth-order regular approximation (ZORA) for relativistic effects as implemented by van Wüllen^24,25. Solvation was modeled with the conductor-like polarizable continuum model (CPCM)²⁶ in a dielectric of 7.25 to simulate a slightly solvent accessible protein active site.

Results and Discussion:

To produce a homogenous solution of Mn-bound S. oleracea Rubisco, natively extracted and purified Rubisco was treated with metal chelator EDTA to sequester Mg(II), inactivating the enzyme. After washing the Mg-EDTA out of solution by gel filtration chromatography, the Rubisco was then treated with a solution containing excess MnCl₂ to reconstitute the active protein and sodium bicarbonate to carbamylate Lys201. RuBP oxygenation activity was assayed via oxygen electrode. Mn-loading was verified via EPR (Figure S3), as well as via oxygen electrode studies (Figure S4). Purified, Mn-loaded Rubisco was then treated with CABP, which inhibited activity as expected (Figure S5). Following Mn-loading and CABP treatment, Rubisco was once again subjected to gel filtration chromatography to remove any excess Mn(II) and CABP. Homogenized and inhibited Mn-bound Rubisco was crystallized via the sitting drop method in the presence of 22% (w/v) polyethylene glycol (MW = ~3350) and 200 mM lithium acetate (pH = 5.3). Diffraction to 2.5 Å was obtained. The space group was found to be P2₁2₁2, with unit cell parameters of a = 218.03, b = 218.89, c = 111.84, α = 90, β = 90, γ = 90 (Table 1). The asymmetric unit comprised the full Rubisco hexadecamer of 8 large and 8 small subunits. The crystal structure was solved using molecular replacement (MR) using a representative Mg(II) and CABP-bound Rubisco structure from spinach (8RUC)⁵ as a starting model.

Table 1:

X-ray diffraction data collection and refinement statistics for Mn-bound S. oleracea Rubisco.

Data collection and merging statistics
Wavelength (Å)	0.984
Temperature (K)	100
Space group	P 21 21 2
a (Å)	218.03
b (Å)	218.89
c (Å)	111.84
α (deg)	90.0
β (deg)	90.0
γ (deg)	90.0
Resolution (Å)	97.89 – 2.5 (2.59 – 2.50)
No. of reflections	2582418 (265389)
Rmerge	0.511 (2.575)
Rmeas	0.531 (2.668)
CC1/2	0.978 (0.574)
Completeness (%)	100.0 (100.0)
Redundancy	14.0 (14.5)
I/σ(I)	8.0 (1.3)
Refinement statistics
Resolution (Å)	99.51 – 2.50 (2.59 – 2.50)
Completeness (%)	100.0 (100.0)
No. of unique reflections	184848 (18299)
No. of unique reflections in the Rwork set	175898 (17411)
No. of unique reflections in the Rfree set	8950 (888)
Rwork (%)	19.65 (29.56)
Rfree (%)	24.24 (34.86)
Root-mean-square deviation from ideality
Bonds (Å)	0.008
Angles (deg)	1.05
Average B factor (Å2)	48.0
Ramachandran Plot
Favored regions (%)	97.30
Allowed regions (%)	2.61
Disallowed regions (%)	0.09
PDBID	9CQ5

The resulting structural model of the active enzyme revealed that the canonical activating modifications to the as-synthesized protein were present in-crystallo, specifically the carbamylated Lys201 and the presence of a divalent metal ion. The electron density map of Lys201 showed expected electron density consistent with carbamylation of the residue and, during the MR process, extra electron density was observed when Mg(II) was placed in the active site, but not when the ion was manually replaced with Mn(II) (Figure 2). Electron density consistent with CABP being bound was also seen in our experiments, as the electron density from the obtained crystallographic data matched the electron density of previously reported structures of Mg-bound Rubisco inhibited by CABP well.

Figure 2:

Comparison of the total density 2F_o–F_C and difference F_o–F_c density maps of chain A models with Mg bound (gold, left) vs Mn bound (purple, right) in refinement. Density maps are plotted at a level of σ = 3 F_o–F_c for and σ = 2 for 2F_o–F_c.

The structure of Mn-bound Rubisco was aligned against the Mg-bound structure to look for structural differences in the active site (Figure 3). Direct comparisons of bond lengths and angles were made against two representative high-resolution Mg-bound structures of Rubisco from spinach (PDB: 8RUC and 1IR1)^5,27 (Table 2). These comparisons show that the inner sphere bond distances are either within crystallographic data error of or between the distances observed for the two representative Mg-bound structures. However, the current understanding of the mechanism of action of Rubisco is that the reactivity is predominantly outer sphere based. Due to the prevailing view of Rubisco’s reactivity as well as our geometric inner sphere results, respective geometric parameters of CABP were also tabulated and compared between our Mn-Rubisco structure and the Mg-bound Rubiscos (Table 2). The C–O bond distances of the Mn-Rubisco show insignificant differences compared against those from the Mg-Rubisco structures, but the C–C bond distances of the CABP in the Mn-Rubisco structure are slightly shorter than in the CABP bound to Mg-Rubisco. Analysis of the bond angles in CABP for Mn- and Mg-Rubisco reveal that there are slight differences in the geometries of the C2 and C3 carbons of CABP, but these differences are minor and, in our opinion, do not readily yield chemical insight regarding the origins of the difference in reactivity between the Mn- and Mg-bound Rubiscos.

Figure 3.

Structural alignment of chain A of Mn-bound Rubisco (green/purple) (PDB: 9CQ5) with chain A of Mg-bound Rubisco (blue/gold) (PDB: 8RUC) RMSD = 0.223 Å.

Table 2.

Geometric parameters obtained from the Mn-bound spinach Rubisco structure and those obtained from two representative Mg-bound spinach Rubisco structures. Values are averaged for each unique metal binding site in the asymmetric unit. Values in parentheses represent the error in the last value estimated using the standard deviation of observed metrics. The labeling scheme is the same as used in Figure 1 and “XLys201” denotes carbamylated Lys 201.

	9CQ5 (Mn)	8RUC (Mg)	1IR1 (Mg)
Distances (Å)
M2+–O2	2.34(8)	2.38(3)	2.35(2)
M2+–O2'	2.20(3)	2.34(2)	2.18(9)
M2+–O3	2.45(9)	2.38(4)	2.26(6)
M2+–XLys201	2.16(8)	2.32(3)	2.12(5)
M2+–Asp203	2.14(8)	2.30(1)	2.04(2)
M2+–Glu204	2.17(6)	2.31(1)	2.05(3)
C2–O2	1.45(1)	1.38(1)	1.44(1)
C2'–O2'	1.24(5)	1.24(1)	1.23(1)
C3–O3	1.41(1)	1.40(1)	1.43(1)
C2–C2'	1.52(0)	1.55(1)	1.58(1)
C2–C3	1.53(4)	1.57(2)	1.57(1)
C3–C4	1.53(0)	1.54(1)	1.56(1)
Angles (°)
∠C1–C2–C2'	108.5(7)	113(1)	112.8(2)
∠C1–C2–C3	110.3(3)	106(2)	109.2(3)
∠C2'–C2–C3	112.7(7)	114(2)	110.0(2)
∠C2–C3–C4	114.2(8)	116(2)	119.6(4)
∠C1–C2–O2	109.9(7)	105(2)	106.8(1)
∠C2'–C2–O2	106.7(7)	109(2)	109.4(1)
∠C2–C2'–O2'	120.4(0)	121.7(9)	119.5(2)
∠C2–C3–O3	110.3(4)	111.4(8)	110.3(1)
∠C3–C2–O2	107.9(6)	109(2)	108.5(3)
∠C4–C3–O3	109.6(7)	104.9(2)	106.4(2)
∠O2'–C2'–O2"	119.0(3)	118.2(1)	118.7(4)

To attempt to further rationalize reactivity differences between Mg- and Mn-bound Rubisco, the crystallographically characterized active sites were used as starting points for density functional theory (DFT) calculations. A minimal model comprising carbamylated Lys201, Asp203, Glu 204, and CABP was excised from each subunit in the asymmetric units of the 8RUC Mg-Rubisco and 9CQ5 Mn-Rubisco structures. Hydrogen atoms were added and optimized using the BP86 GGA density functional. The resulting models were then used as input structures for hybrid DFT calculations using the B3LYP functional and the all-electron, scalar relativistically recontracted ZORA-def2-TZVP(-f) basis set. Lowdin charge analysis was carried out to determine whether the potential for bonding with a greater degree of covalency between CABP and Mn resulted in significant perturbation to this bound molecule relative to the more expected ionic bonding in the Mg case. Lowdin atomic charges^28,29 for atoms of interest were averaged for all unique metal sites in the asymmetric unit of the crystal structure. These charges are collected in Table 3. Charges at the bound CABP are effectively invariant between the structures. We assume that a similar situation would be encountered with bound RuBP, for which we do not have experimental structural data as a starting point for calculations.

Table 3.

B3LYP/ZORA-def2-TZVP(-f)-calculated Lowdin atomic charges on active site metal and select CABP atoms in Mn- and Mg-bound Rubisco active site models. Charges are given in units of e and represent the averages of all unique metal sites in the asymmetric unit, and the final parenthetical values represent estimated error based on the standard deviation of these values.

Atom	9CQ5 (Mn)	8RUC (Mg)
M	0.28(6)	.264(1)
C1	−.221(8)	−.201(3)
C2	−.232(9)	−.26(1)
C2'	−.42(2)	−.44(1)
C3	−.204(8)	−.196(8)
C4	−.249(6)	−.273(9)
O2	0.03(3)	0.076(3)
O2'	−.05(4)	−.048(9)
O3	0.03(3)	.059(5)

Conclusions:

The structure of a Mn-bound Rubisco was obtained to facilitate comparisons to the Mg-bound form. Lys carbamylation was structurally confirmed for the Mn-bound form of the enzyme. Negligible differences in bond length and bond angles were observed between the Mg- and Mn-bound structures, suggesting that this difference in activity is not primarily due to a structural difference in the active site. DFT calculations carried out to compare atomic charges on bound substrate analogue also reveal insignificant differences between Mg- and Mn-bound Rubisco forms. Together, the experimental data and computational results do not readily yield an explanation for the observed differences between the relative rates of reaction of Mg- and Mn-bound Rubisco with O₂ and CO₂. Regardless, the presently reported structure should serve as a useful starting point for further investigation. For example, dynamical properties rather than static, ground state properties may play a greater role in biasing the different metal-bound forms of Rubisco between CO₂ fixation and RuBP oxygenation. The Mn-Rubisco structure reported here confirms that major structural perturbations need not be a factor at the outset of such studies.

Highlights.

• Crystal structure of Rubisco with the alternative physiological metallocofactor Mn2+.

• Confirmation of activation with bound alternative metallocofactor.

• Density functional theory comparison of Mg2+ and Mn2+-bound variants