Archives of Biochemistry and Biophysics 536 (2013) 87–96

Contents lists available at SciVerse ScienceDirect

Archives of Biochemistry and Biophysics

journal homepage: www.elsevier .com/ locate/yabbi

Folate polyglutamylation eliminates dependence of activity on enzymeconcentration in mitochondrial serine hydroxymethyltransferasesfrom Arabidopsis thaliana

0003-9861/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.abb.2013.06.004

⇑ Corresponding author. Fax: +1 509 335 7643.E-mail address: sanja@wsu.edu (S. Roje).

1 Present address: Biology Department, Brookhaven National Laboratory, Upton, NY11973, USA.

2 Abbreviations used: (6S)-5-CH3–H4PteGlun, 5-methyltetrahydrofolate; (6S)-5,10-CH@H4PteGlun, 5,10-methenyltetrahydrofolate; (6S)-5,10-CH2–H4PteGlun, 5,10-methylenetetrahydrofolate; (6S)-10-HCO–H4PteGlun, 10-formyltetrahydrofolate;(6S)-H4PteGlun, tetrahydrofolate; EGFP, enhanced green fluorescent protein; Ni–NTA, nickel–nitrilotriacetic acid; PLP, pyridoxal 50-phosphate; SHMT, serine hydrox-ymethyltransferase; THP, tris-(3-hydroxypropyl)phosphine.

Zhaoyang Wei a,1, Kehan Sun a, Francisco J. Sandoval a, Joanna M. Cross b, Christine Gordon c,ChulHee Kang d, Sanja Roje a,⇑a Institute of Biological Chemistry, Washington State University, Pullman, WA 99164, USAb German Institute of Food Technologies, Prof.-von-Klitzing-Str. 7, D-49610 Quakenbrück, Germanyc Pacific Lutheran University, Tacoma, WA 98447, USAd Department of Chemistry, Washington State University, Pullman, WA 99164, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 13 March 2013and in revised form 11 June 2013Available online 22 June 2013

Keywords:One-carbon metabolismFolate polyglutamylationSerine hydroxymethyltransferase (SHMT)Tetrahydrofolate

The reversible reaction catalyzed by serine hydroxymethyltransferase (SHMT) is the major one-carbonunit source for essential metabolic processes. The Arabidopsis thaliana genome encodes seven SHMT iso-zymes localized in mitochondria, plastids, nuclei, and the cytosol. Knowledge of the biochemical proper-ties of each isozyme is central to understanding and manipulating one-carbon metabolism in plants. Weheterologously expressed and purified three recombinant SHMTs from A. thaliana (AtSHMTs) putativelylocalized in mitochondria (two) and the cytosol (one). Their biochemical properties were characterizedwith respect to the impact of folate polyglutamylation on substrate saturation kinetics. The two mito-chondrial AtSHMTs, but not the cytosolic one, had increased turnover rates at higher (>0.4 ng/lL) enzymeconcentrations in the presence of monoglutamylated folate substrates, but not in the presence of penta-glutamylated folate substrates. We found no experimental support for a change in oligomerization stateover the range of enzyme concentration studied. Modeling of the enzyme structures presented featuresthat may explain the activity differences between the mitochondrial and cytosolic isozymes.

� 2013 Elsevier Inc. All rights reserved.

Introduction

Serine hydroxymethyltransferase (SHMT; EC 2.1.2.1)2 catalyzesthe reversible reaction L-serine + (6S)-H4PteGlun (Fig. 1) ? gly-cine + (6S)-5,10-CH2–H4PteGlun [1–5]. In plants, SHMT activity wasdetected in mitochondria [2,6–9], plastids [2,7,10], the cytosol[2,6,7], and nuclei [7]. The Arabidopsis thaliana genome harbors se-ven genes (AtSHM1–7) encoding SHMT isozymes putatively localizedin these subcellular compartments [11]. Subcellular localization wasconfirmed for the isozymes from mitochondria (AtSHM1 and 2)[12,13] and plastids (AtSHM3) [10], but remains to be confirmed

for those putatively localized in the cytosol (AtSHM4 and 5) and nu-clei (AtSHM6 and 7).

In plastids and the cytosol [2,14–16], SHMTs provide one-car-bon units to the cellular folate pool by producing (6S)-5,10-CH2–H4PteGlun, which is then reduced to (6S)-5-CH3–H4PteGlun by5,10-CH2–H4PteGlun reductase (EC 1.5.1.20) or oxidized sequen-tially by 5,10-CH2–H4PteGlun dehydrogenase (EC 1.5.1.5) and5,10-CH@H4PteGlun cyclohydrolase (EC 3.5.4.9), respectively yield-ing (6S)-5,10-CH@H4PteGlun and (6S)-10-HCO–H4PteGlun [17].These folate derivatives are essential for nucleotide and amino acidbiosyntheses, methyl group biogenesis, and vitamin metabolism[14,17,18]. During photorespiration in mitochondria of C3 plantcells [16,19,20], SHMTs act in concert with the glycine decarboxyl-ase complex (EC 1.4.4.2) to convert two molecules of glycineformed during photorespiration into one molecule of serine[11,18,21]. A functional photorespiratory pathway is essential toplants, and mutants in this pathway have to be grown under ele-vated CO2 concentrations to suppress photorespiration [22]. A con-ditional lethal mutant of A. thaliana, shm1–1, is deficient in themitochondrial AtSHMT1 [23].

In plants [24,25], as in other organisms [26], polyglutamylatedspecies dominate cellular folate pools. Penta- and hexaglutamylat-

Fig. 1. Chemical structure of H4PteGlun.

88 Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96

ed folates, with tetra- and heptaglutamylated forms following, arethe most abundant species in most eukaryotic cells [27]. Polygluta-mylated folates comprise over 80% of the folate pool of whole cells,as well as of mitochondria and chloroplasts, in A. thaliana leaves[25]. Polyglutamylated species also dominate the matrix spacepool of pea leaf mitochondria, with over 50% of folates having pen-taglutamyl or longer tails [24].

Previous studies have shown that polyglutamylation yields effi-cient substrates for some folate-dependent enzymes by favoringprotein binding [28,29]. Dependence of activity on folate polyglut-amylation has been investigated in few plant enzymes. The A. tha-liana methionine synthase shows a higher preference forpolyglutamylated over monoglutamylated 5-CH3–H4PteGlun

[30,31]. The glycine decarboxylase complex and the SHMT frompea leaf mitochondria have lower Km values for polyglutamylatedthan for monoglutamylated H4PteGlun [24,32]; the catalytic con-stants for the H4PteGlun > 2 could not be determined for the SHMTbecause of limitations in assay sensitivity. We previously demon-strated that polyglutamylated folates are better substrates and bet-ter inhibitors than the monoglutamylated species for the reactionin the direction of glycine formation for the plastid AtSHMT3 fromA. thaliana [10]. Catalytic properties remain to be determined forthe remaining AtSHMT isozymes.

Here we report cDNA cloning, recombinant expression, purifica-tion, and biochemical characterization of the mitochondrial AtS-HMT1 and AtSHMT2 [11,23,33], and the cytosolic AtSHMT4[11,33,34]. The second putative cytosolic isozyme, AtSHMT5, wasalso expressed, but the purified protein was catalytically inactiveand, therefore, not studied further.

Subcellular localization of AtSHMT1 and AtSHMT2 has beenconfirmed [12,13]. We here provide evidence that AtSHMT4 re-sides in the cytosol using fluorescence microscopy of the enzymefused to enhanced green fluorescent protein (EGFP). Catalytic prop-erties of AtSHMT1, AtSHMT2, and AtSHMT4 were studied using anHPLC-based fluorometric assay we developed previously [36]. AtS-HMT1 and AtSHMT2, but not AtSHMT4, displayed an increase incatalytic activity at higher (>0.4 ng/lL) enzyme concentrationwhen assayed in the presence of monoglutamylated folate sub-strates, but not in the presence of pentaglutamylated folate sub-strates. Additional experiments did not support the hypothesisthat the increased activity at higher enzyme concentration is dueto increased dimer-to-tetramer conversion in the enzymes frommitochondria. Modeling of the enzyme structures presented fea-tures that may explain the differences in the kinetic parametersbetween the mitochondrial and cytosolic isozymes.

Materials and methods

Chemicals and reagents

PteGlu5, (6R,S)-H4PteGlu1, and (6R,S)-5,10-CH2–H4PteGlu1 wereobtained from Schircks Laboratories (Jona, Switzerland). NaBH4

was from Sigma–Aldrich. Glutaraldehyde (25% aqueous solution)was from Fisher Scientific. BSA was from EMD Chemicals (Darms-

tadt, Germany). Benzonase� nuclease, recombinant enterokinase,and Ni–NTA His�Bind� superflow resin were from Novagen (Mad-ison, WI). Oligonucleotides were from MWG (High Point, NC). Therabbit polyclonal anti-Spinacia oleracea SHMT antibody was fromAgrisera (Vännäs, Sweden).

cDNA cloning and expression in E. coli

The corresponding cDNA clones were ordered from the Arabid-opsis Biological Resource Center (ABCR), The Ohio State University,for AtSHMT1 (At4g37930, ABRC clone 135G2), AtSHMT2 (At5g26780,ABRC clone C104687), AtSHMT4 (At4g13930, ABRC clone F1D1T7),and AtSHMT5 (At4g13890, ABRC clone 160C13). For AtSHMT2,two protein sequences are present, AtSHMT2 and AtSHMT2long,the second differing from the first in having a 16-amino acid inser-tion (Fig. S1). The ABRC clone C104687 encodes AtSHMT2long.

The cDNA sequences encoding the N-terminal regions of AtS-HMT1 and AtSHMT2 or AtSHMT2long were recoded by DNA syn-thesis using a commercial service (GenScript, Piscataway, NJ) tocomply with the codon usage bias in E. coli and facilitate expres-sion. The full-length coding sequences were then reconstitutedthrough overlap extension PCR [37]. First, the synthesized recodedsequence of AtSHMT1 (300 bp) was amplified by PCR using the pri-mer pair ForwardA and ReverseA (Table S1). The remaining AtS-HMT1 sequence was amplified from clone 135G2 using primersForwardB (complementary to ReverseA) and ReverseB (Table S1).The full-length open reading frame of AtSHMT1 was producedusing the two amplified sequences as templates and the primerpair AtSHM1 F and R (Table S1). The resulting PCR fragments werepurified with Wizard PCR prep mini-columns (Promega, Madison,WI), treated with T4 DNA polymerase (Promega), and then insertedinto the pET-43.1 Ek/LIC vector (Novagen). All procedures werecarried out in accordance with the manufacturer’s protocols. Pro-teins expressed in pET-43.1 Ek/LIC vectors have an N-terminalNus�Tag fusion partner, which increases solubility of the recombi-nant proteins.

The coding sequence for the N-terminal region of AtSHMT2 orAtSHMT2long (300 bp) was amplified using the primer pair For-wardC and ReverseC (Table S1). The coding sequence for the C-ter-minal region of AtSHMT2long was amplified from clone C104687using primers ForwardD and ReverseD (Table S1). The full-lengthrecoded open reading frame of AtSHMT2long was produced usingthe two PCR products as templates and the primer pair AtSHM2F and R (Table S1).

The coding sequence for the C-terminal region of AtSHMT2 wasamplified through another overlap extension PCR to eliminate thesequence encoding the 16-amino acid insertion. The cDNA se-quences comprising the segment were amplified from cloneC104687 using primer pairs ForwardD and ReverseE, and ForwardFand ReverseD (Table S1). Primers ReverseE and ForwardF(Table S1), which are complementary to each other, introducethe deletion required to reconstitute AtSHMT2. The full-length cod-ing sequence for AtSHMT2 was then amplified using the primerpair AtSHM2 F and R (Table S1), using as templates the PCR prod-ucts encoding the N- and C-terminal regions of AtSHMT2. The full-length coding sequences for AtSHMT2 and AtSHMT2long werepurified and inserted into the pET-30 Ek/LIC vector following themanufacturer’s protocol. PCR applications used PfuTurbo DNA poly-merase (Stratagene, La Jolla, CA).

The F1D1T7 clone containing the AtSHMT4 open reading frameharbored a four-base pair deletion in the middle of the sequence.The deletion was corrected also using overlap extension PCR. Thetwo segments flanking the deletion were amplified respectivelywith primer pairs ForwardG and ReverseG, and ForwardH and Re-verseH (Table S1). Primers ReverseG and ForwardH introduced thefour-base pair insertion. Next, the two PCR products were ampli-

Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96 89

fied with the primer pair AtSHM4 F and R (Table S1) to produce thefull-length coding sequence, adding an NdeI site at the 50 end andan XhoI site at the 30 end. The coding sequence for AtSHMT5 wasamplified using the primer pair AtSHM5 F and R (Table S1), addingan NdeI site at the 50 end and an XhoI site at the 30 end. PCR appli-cations used PfuTurbo DNA polymerase. The amplified coding se-quences for AtSHMT4 and AtSHMT5 were digested with NdeI andXhoI and then inserted into the pET-43.1b(+) vector also digestedwith those enzymes. Thus, AtSHMT4 and AtSHMT5 were expressedwithout any fusion tag.

The generated expression vectors were introduced into E. coliNovaBlue competent cells (Novagen), sequence-verified, and thenintroduced into E. coli Rosetta 2(DE3) (Novagen) for protein produc-tion. To express the recombinant proteins, bacteria carrying theexpression vectors were incubated at 37 �C in Luria–Bertani med-ium containing 50 lg/mL chloramphenicol and 50 lg/mL ampicillinor 100 lg/mL kanamycin until the absorbance at 600 nm reached0.6–1.0. Isopropyl b-D-thiogalactopyranoside (200 lM was thenadded, and incubation was continued at 15 �C overnight.

Protein purification

Induced E. coli cells from 200 mL cultures were harvested by cen-trifugation at 7500g for 10 min at 4 �C, resuspended in 5 mL of bufferA (50 mM CHES-HEPES-citric acid buffer (pH 7.5) [38], 1 mM tris-(3-hydroxypropyl)phosphine (THP), 0.25 mM pyridoxal 50-phosphate(PLP), and 10% glycerol) plus 25 units/mL Benzonase� nuclease,and then lysed with 0.1 mm zirconia/silica beads using a Mini-Bead-beater-8 cell disrupter (BioSpec Products, Bartlesville, OK). Solubleprotein extracts were cleared by centrifugation at 20800g for15 min at 4 �C and filtered through a 0.45-lm PVDF membrane.

Recombinant AtSHMT1, AtSHMT2, and AtSHMT2long wereaffinity-purified on Ni–NTA His�Bind� resin (Novagen) as in-structed by the manufacturer. Protein preparations were incubatedwith 1 mL of Ni–NTA resin for 1 h at 8 �C. Binding, wash, and elu-tion buffers contained 50 mM CHES–HEPES–citric acid buffer (pH7.5), 300 mM NaCl, 10 (binding), 20 (wash), or 150 mM (elution)imidazole, 0.5 mM THP, 0.25 mM PLP, and 10% glycerol. The puri-fied proteins were immediately desalted into buffer A with PD-10 desalting columns (GE Healthcare) and digested with recombi-nant enterokinase to cleave the fusion tags. The best reaction con-ditions for enterokinase cleavage were checked by SDS–PAGE usingthe NuPAGE Novex� Bis-Tris gel system (Invitrogen, Carlsbad, CA).Protein digests were stained with the SimplyBlue SafeStain� (Invit-rogen). The fully cleaved enzymes were further purified to removethe tags and undigested enzymes using the ÄKTA FPLC system (GEHealthcare). Untagged AtSHMT1 was purified into buffer A on aMono Q 5/50 GL column (GE Healthcare) at 0.75 mL min–1. Thepurified protein eluted in the flow-through. Untagged AtSHMT2was purified into buffer A on a Superose 12 10/300 GL column(GE Healthcare) at 0.8 mL min–1. The volume of sample injectedwas 5 mL. The tagged SHMT2 long was used for characterizationbecause of the inefficiency of tag cleavage for this enzyme.

Recombinant AtSHMT4 and AtSHMT5 were purified using threechromatographic steps. First, protein preparations were loadedonto a HiPrep 16/10 DEAE FF column (GE Healthcare) and elutedwith a gradient of 0–0.5 M NaCl in buffer A over 5 column volumesat 1 mL min–1. The purified proteins eluted at �0.1 M NaCl. Second,the active fractions were loaded onto a Mono Q 5/50 GL columnand eluted with a gradient of 0–0.5 M NaCl in buffer A over 10 col-umn volumes at 0.75 mL min–1. The purified proteins eluted at�0.25 M NaCl. Third, the active fractions were purified into bufferA on a Superdex 200 10/300 GL column (GE Healthcare) at0.5 mL min–1. The volume of sample injected was 5 mL. All purifi-cation steps were checked by SDS–PAGE. The purified enzymeswere frozen in liquid N2 and stored at �80 �C until use.

Synthesis of H4PteGlu5

PteGlu5 was reduced to H4PteGlu5 with NaBH4 [25] for use inthe activity assays. The resulting H4PteGlu5 was purified as de-scribed previously [10]. H4PteGlu5 concentration was determinedspectrophotometrically at 298 nm with a molar absorption coeffi-cient of 28,400 M–1 cm–1. The purified H4PteGlu5 solution wasstored in tightly sealed vials at –80 �C.

SHMT activity assay

SHMT activity was measured using an HPLC-based fluorometricassay [36], with modifications as indicated below. Reaction prod-ucts were analyzed using an Alliance 2695 separations modulecoupled with a 2475 fluorescence detector (Waters, Milford, MA).The enzymatic reaction contained 50 mM CHES–HEPES–citric acid(pH 7.5), 4 mM THP, 0.25 mM PLP, and 100 ng/lL BSA plus 5 mM L-serine and various (6R,S)-H4PteGlu1&5 concentrations. The reactionvolume was 50 lL. Assay mixtures were pre-incubated on ice for10 min before addition of serine to start the reaction. Pre-incuba-tion for 30 min had no effect on SHMT activity (not shown). Serinewas omitted solely from the assay blanks before incubation. BSAwas included in the reaction mixture because preliminary results(not shown) established that SHMT activity increases 5–6-fold inthe presence of BSA in all three enzymes.

After incubation at 22 �C for 20 min, the reactions were stoppedby simultaneously adding 25 lL of 0.1 M dithiothreitol, which low-ers blank values, and 50 lL of 0.1 M NaBH4. The NaBH4 solutionwas freshly prepared before use because it degrades quickly inwater. The reaction products were incubated at 37 �C for 15 minto drive (6S)-5-CH3–H4PteGlu1&5 formation to completion, heatedat 98 �C for 3 min, and centrifuged at 2000g for 15 min at 4 �C topellet denatured proteins. To prevent product decomposition,10 lL of 0.6 M dithiothreitol per 50 lL of supernatant was addedto the reaction products. (6S)-5-CH3–H4PteGlu1&5 and (6R,S)-H4-

PteGlu1&5 were separated isocratically on a Waters XTerra C18 col-umn (4.6 � 100 mm, 5 lm) and detected fluorometrically at280 nm excitation and 359 nm emission wavelengths. The mobilephase consisted of 27 mM phosphoric acid and 7 or 9% (v/v) meth-anol when using mono- or pentaglutamylated folates, respectively.Flow conditions during the isocratic separation were 2 mL min–1

for 6–12 min at 32 �C. The volume of sample injected was 2–20 lL. The (6S)-5-CH3–H4PteGlu1&5 formed during the assays werequantified by comparison with standards. Reaction products in-creased linearly with time and enzyme concentration in all assayconditions used here.Apparent values for the kinetic parameterswere found by fitting initial reaction rates against substrate con-centrations to a model of uncompetitive substrate inhibition inthe Enzyme Kinetics module 1.2 of SigmaPlot 9.0.1 (Systat Soft-ware, San Jose, CA). The rate equation used to fit the data isv = Vmax/(1 + Km/S + S/Ki) in which v is the initial reaction rate, Vmax

is the limiting rate, Km is the Michaelis constant, S is the initial sub-strate concentration, and Ki is the inhibition constant. To ease dataprocessing and plotting, the reaction rates from three independentassays, each carried out in duplicate, were grouped into a singledataset and simultaneously entered into SigmaPlot 9.0.1 to providethe reported kinetic parameters and standard errors. The standarderror for kcat/Km and Ki/Km was calculated by error propagation.

Protein cross-linking and Western blot analysis

A 10–15 lL sample of SHMT at concentrations corresponding tothe range used in the activity assays was cross-linked with glutar-aldehyde without direct mixing as described previously [40]. Glu-taraldehyde is a volatile liquid, which forms vapor pervading theincubation cell that dissolves in the sample drops. This is a mild

90 Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96

method not requiring direct mixing of the sample with the cross-linker. Proteins were diluted in a buffer solution containing50 mM CHES–HEPES–citric acid (pH 7.5), 4 mM THP, and0.25 mM PLP plus 534 lM (6R,S)-H4PteGlu1. The cross-linked pro-teins were separated by SDS–PAGE using the NativePAGE Novex�

Bis-Tris gel system (Invitrogen) and analyzed by Western blot withchromogenic detection using the amplified alkaline phosphataseImmun-Blot� assay kit (Bio-Rad, Hercules, CA). The rabbit poly-clonal anti-Spinacia oleracea SHMT antibody was diluted 5000times before use.

Structural modeling of AtSHMT1, AtSHMT2, and AtSHMT4

The crystal structure of a human cytosolic SHMT with boundpyridoxamine 50-phosphate (PDB ID: 1BJ4) [41] was used to builda molecular model for AtSHMT1, AtSHMT2, and AtSHMT4. Theamino acid substitutions were conducted using the coordinatesof the human cytosolic SHMT and then regularized using definedgeometric constraints for bonds, angles, planes, non-bonded con-tacts, and torsion constraints in Coot [42]. After geometric regular-ization, quick global energy minimization was carried out usingCNS 1.1 [43], which uses the potential function parameters ofCHARMM19. Substrate position was generated using the soliddocking module on QUANTA (BioSYM/Micron Separations), whichis based on conformational space, followed by a quick energy min-imization by CNS 1.1.

Transient expression of EGFP-fused AtSHMT4 in A. thaliana protoplasts

The full-length coding sequence for AtSHMT4 was amplified byPCR with forward primer AtSHMT4 GFPF and reverse primers AtS-HMT4 C-terminus GFPR or AtSHMT4 N-terminus GFPR (Table S1)for C- or N-terminal EGFP fusion, respectively. The resulting plas-mid fragments were re-amplified with the attB adapter primer pair(Invitrogen), purified with Wizard PCR prep mini-columns, andthen inserted into the p2GWF7 or p2FGW7 vector for C- or N-ter-minal EGFP fusion, respectively. All procedures were carried out inaccordance with the Gateway cloning system (Invitrogen).

Arabidopsis thaliana protoplasts were isolated from leaves of 4-week-old plants and transformed using a polyethylene glycol-med-iated transformation method [44] for transient gene expression.Fluorescence was monitored using a TCS SP5 confocal laser-scan-ning microscope (Leica Microsystems, Exton, PA). EGFP fluorescencewas excited at 488 nm and detected at 505–530 nm. Chlorophyllfluorescence was excited at 488 nm and detected at >650 nm.

Results

Sequence analysis

Deduced amino acid sequences for AtSHMT1, AtSHMT2, AtS-HMT4, and AtSHMT5 were retrieved from The Arabidopsis Informa-tion Resource database. Three gene models for AtSHMT2 werepresent. Two of them, At5g26780.2 and At5g26780.3, encoded thesame putative protein (AtSHMT2long) and one, At5g26780.1, en-coded a sequence variant (AtSHMT2). AtSHMT2long has a 16-aminoacid insertion (amino acids 373–388) relative to AtSHMT2 (Fig. S1).WoLF PSORT prediction suggested that AtSHMT1 and AtSHMT2 haveN-terminal extensions for targeting to mitochondria, while AtS-HMT4 and AtSHMT5 have no such extensions (Fig. S1) and thusare predicted to be cytosolic. Previously, mitochondrial localizationof AtSHMT1 and AtSHMT2 was confirmed respectively by fluores-cence microscopy of stable A. thaliana transformants expressing aC-terminal AtSHMT1-GFP fusion [12] and by Western blotting ofmatrix extracts prepared from purified mitochondria [13].

Purification of recombinant AtSHMTs

Recombinant AtSHMT1, AtSHMT2, and AtSHMT2long were ex-pressed in E. coli from pET-30 or pET-43.1 Ek/LIC. The tag was suc-cessfully cleaved from AtSHMT1 and AtSHMT2, but not fromAtSHMT2long (Fig. 2). Because AtSHMT2 activity did not changeafter the tag cleavage (not shown), we used the tagged AtS-HMT2long for the activity assays. Preliminary results (not shown)indicated that the tags could not be cleaved from AtSHMT4 or AtS-HMT5 expressed from pET-30 and pET-44 Ek/LIC. These two en-zymes were therefore expressed and purified without any tags(Fig. 2).

Kinetic characterization of AtSHMTs

AtSHMT1, AtSHMT2, and AtSHMT4 were catalytically active andthus assayed for activity in the presence of mono- or pentaglutam-ylated folate substrates. Specific activity increased with enzymeconcentration in AtSHMT1 and AtSHMT2 when assayed in thepresence of monoglutamylated folates (Fig. 3A and C), but not inthe presence of pentaglutamylated folates (Fig. 3B and D). Specificactivity of AtSHMT4 remained constant with enzyme concentra-tion in the presence of either monoglutamylated (Fig. 3E) or penta-glutamylated (Fig. 3F) folates. Preliminary results (not shown)established that the effect of enzyme concentration on specificactivity persisted when BSA was omitted from the assays. Anexplanation for the observed dependence of specific activity on en-zyme concentration for AtSHMT1 and AtSHMT2 is offered inDiscussion.

Kinetic parameters derived from nonlinear data fitting to amodel of uncompetitive substrate inhibition (SigmaPlot 9.0.1) arepresented in Table 1. Substrate saturation curves used to calculatethe kinetic parameters are shown in Fig. S2. The kcat value for AtS-HMT1 and AtSHMT2 could not be determined because of twoexperimental restrictions. First, the activity of these two enzymeschanged with enzyme concentration; second, the enzyme concen-tration would have to be changed when assaying all substrate con-centrations to control substrate consumption and the signal-to-blank ratio. The Km value for AtSHMT1 and AtSHMT2 could bedetermined only if the enzyme concentration would be held con-stant across the monoglutamylated folate concentrations assayed.This was not feasible because the enzyme concentration was eithertoo high for the lowest substrate concentration points, so morethan 10% of substrate would be consumed and the rate of the reac-tion would move too far away from the steady state, or too low forthe highest substrate concentration points, requiring past the 50-min incubation time. Our preliminary results showed that thereaction product increases linearly with time for up to 50 min.

The data showed for all three enzymes that the Km values arelower and the catalytic efficiency (kcat/Km) is higher in the presenceof pentaglutamylated folates than in the presence of monoglutam-ylated folates. Substrate inhibition was also more pronounced forthe polyglutamylated substrates. Ki of AtSHMT4 is �190-fold lowerfor H4PteGlu5 than for H4PteGlu1. The decrease in Ki is relativelylarger than the decrease in Km, resulting in a �5-fold decrease inthe Ki/Km value. This is different from the previously studied AtS-HMT3, for which Ki decreases to a lesser degree than Km with theincrease in the level of folate polyglutamylation [10].

Oligomerization state of AtSHMT1 and AtSHMT2

Previous studies have shown that SHMTs exist as homotetra-mers best described as dimers of dimers in eukaryotes such as ze-bra fish, mouse, rabbit, sheep, and human [41,45–48] and asdimers in prokaryotes such as E. coli, Bacillus subtilis, and Salmo-nella typhimurium [5,49,50]. However, a dynamic balance between

Fig. 2. SDS–PAGE of purified AtSHMT1, AtSHMT2, AtSHMT2long, AtSHMT4, and AtSHMT5. The recombinant enzymes were heterologously expressed in E. coli. AtSHMT1,AtSHMT2, and AtSHMT2long were purified by Ni-NTA affinity chromatography under native conditions. Tagged AtSHMT1 and AtSHMT2 (lanes 1) were cleaved withrecombinant enterokinase and the untagged proteins (lanes 2) were purified respectively on Mono Q and Superose 12 columns. Recombinant AtSHMT4 and AtSHMT5 werepurified using the three chromatographic steps described in Materials and Methods. The purified proteins were separated by SDS-PAGE on NuPAGE Novex� 4–12% Bis–Trisgels and stained with Coomassie Blue. Protein markers are indicated (in kDa) next to the purified proteins.

Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96 91

dimeric and tetrameric forms of some SHMTs has also been re-ported, and catalytic activity changed with the oligomerizationstate in these enzymes [51–53]. Therefore, we hypothesized thata concentration-dependent conversion between dimeric and tetra-meric forms may be causing the increase in specific activity of AtS-HMT1 and AtSHMT2 at higher enzyme concentrations (Fig. 2).

The oligomerization state of AtSHMT1 and AtSHMT2 was stud-ied using mild glutaraldehyde cross-linking in which proteins werenot directly mixed with the glutaraldehyde solution [40]. AtSHMT1and AtSHMT2 were cross-linked at various enzyme concentrationsover the range in which the catalytic activity changed. Since theenzyme quantity was too low to visualize by directly stainingSDS–PAGE gels, the cross-linked proteins were visualized usingantibodies.

The proteins were first cross-linked for 40 min. This reactiontime is sufficient for the cross-linking reaction to reach completion.Since the glutaraldehyde cross-linking is an irreversible processthat perturbs the equilibrium of different oligomerization states,we checked whether the 40-min cross-linking might have pulledthe equilibrium too far toward tetramers. Thus, the proteins werealso partially cross-linked for 30 min.

Western blot analysis of partially cross-linked proteins, in com-bination with the analysis of the band density using MYImageAnal-ysis software (Pierce, Rockford, IL) suggested no transition fromdimer to tetramer with an increase in enzyme concentration(Fig. 4B and C); the tetramer prevailed and the dimer was barelydetectable under the conditions tested (Fig. 4A–C). Therefore, ourexperimental data do not support the hypothesis that the effectof enzyme concentration on specific activity is due to changes inthe oligomerization state of these proteins. Non-denaturing PAGEconfirmed that AtSHMT1 and AtSHMT2 exist as �250-kDa homo-tetramers when the enzyme concentration is high (Fig. 4D). Other

methods to explore oligomerization state of SHMTs, namely dy-namic/static light scattering, fluorescence polarization, molecularsieve chromatography, and analytical centrifugation could not beused because of method limitations detailed in Discussion.

Structural modeling of AtSHMT1, AtSHMT2, and AtSHMT4

Given the high sequence similarity between the human [41]and A. thaliana SHMTs, a reliable 3D-model could be built(Fig. 5A) and served as a foundation to explain the differentialkinetics of AtSHMT1, AtSHMT2, and AtSHMT4. Most of the partic-ipating residues for binding of PLP and pteridine and for catalysisin AtSHMT1, AtSHMT2, and AtSHMT4 were nearly identical(Fig. 5B). Completely conserved among the three isozymes arethe residues for formation of internal aldimine and for ion pairingwith PLP (Lys286 and Asp257), the residues interacting with exter-nal aldimine (Tyr112, His260, and Arg430), the residues interactingwith 5’-phosphate of PLP (Tyr102 and His285), and the residuesinteracting with folate (Asn415 and Tyr111) (Fig. 5B). However,the entry site for folate substrates, which is formed by five loopsfrom the two interfacing subunits, showed substantial heterogene-ity between the mitochondrial and cytosolic isozymes, while thesequences of AtSHMT1 and AtSHMT2 were identical in this region(Fig. 5B).

In general, the entry sites of AtSHMT1 and AtSHMT2 had morerigidity than that of AtSHMT4. The side chain of Asp185, locating atthe tip of the highly disordered loop (173PHGGHLSHGYQTDTKKI-SAVSI194) in both AtSHMT1 and AtSHMT2, formed a salt bridgewith Lys251 in another loop constituting the folate entry point(Fig. 5A). The D2 symmetry of the tetramer enabled four such ionicinteractions, which could introduce significant rigidity to the entrysites and stability to the quaternary structure of AtSHMT1 and AtS-

Fig. 3. Dependence of activity on enzyme concentration for AtSHMT1, AtSHMT2, and AtSHMT4. Activities were assayed at pH 7.5 using purified enzymes under nativeconditions. (A), (C), and (E), when assayed with 267 lM (6S)-H4PteGlu1. (B), (D), and (F), when assayed with saturating (6S)-H4PteGlu5, which is at 50 lM in (B) and 25 lM in(D) and (F). Results are means ± S.E.M. of three independent assays, each carried out in duplicate.

Table 1Kinetic parameters of the AtSHMTs for mono- and pentaglutamylated folate substrates. Kinetic parameters are nonlinear best fits to a model of uncompetitive substrate inhibition(SigmaPlot 9.0.1). The enzymes were assayed for activity with H4PteGlu1 (A) and H4PteGlu5 (B) substrates. In (A), the kinetic parameters for AtSHMT1 and AtSHMT2 are notavailable due to the dependence of activity on enzyme concentration. Results are means ± SEM of three independent assays, each carried out in duplicate.

Enzyme H4PteGlu1

Km (lM) Ki (mM) kcat (s�1) kcat/Km (s�1 lM�1) Ki/Km

A

AtSHMT4 17 ± 4 0.74 ± 0.3 4.6 ± 0.4 0.3 ± 0.1 43 ± 19B

AtSHMT1 28.9 ± 0.9 24 ± 11 10.2 ± 3.3 0.4 ± 0.2 0.8 ± 0.5AtSHMT2 3.0 ± 0.6 10 ± 2 4.5 ± 0.6 1.5 ± 0.4 3.2 ± 1.1AtSHMT4 0.5 ± 0.1 4 ± 1 8.3 ± 1.0 15.9 ± 4.4 8.2 ± 2.7

92 Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96

HMT2. The polyglutamyl moiety of the folate substrates mightweaken the Asp185-Lys251 salt bridge or might interact with thepolar side chains of the entry site, either of which could influencesubstrate entrance/exit or binding.

In AtSHMT4 the corresponding residue for Asp185 is serine andfor Lys251 is glycine, thereby disabling the corresponding saltbridge (Fig. 5B). Also, AtSHMT4 has an inserted glycine after theserine residue (Fig. 5B, dotted box b) forming consecutive glycinesin this region. Three implications are clear from the presented re-sults. First, both loops are more disordered and flexible in this en-zyme. Second, the peptide backbone of the N-terminal residues(Fig. 5B, dotted box b) accommodates part of the pocket for pteri-dine binding and thus the effect of consecutive glycines couldpropagate into this pocket. Third, the constituting residues forthe entry site of AtSHMT4 are less polar than those of AtSHMT1

and AtSHMT2 (Fig. 5B). Overall, the size, flexibility, and polarityof the folate entry site are clearly different between the mitochon-drial and cytosolic isozymes. It is therefore probable that AtSHMT4shows the better turnover rate as a result of this flexibility.

Transient expression of EGFP-fused AtSHMT4 in A. thaliana protoplasts

Mitochondrial localization of AtSHMT1 and AtSHMT2 was con-firmed previously [12,13]. AtSHMT4 was found to localize in thecytosol [54] or plasma membrane [55,56] by proteomic ap-proaches. We confirmed its cytosolic localization using fluores-cence microscopy of the EGFP-fused protein. Transientexpression of both the C- and N-terminal fusions of AtSHMT4 toEGFP produced green fluorescence in the cytosol of the trans-formed A. thaliana protoplasts (Fig. 6). A construct expressing EGFP

Fig. 4. Oligomerization state of AtSHMT1 and AtSHMT2. The oligomerization state of AtSHMT1 and AtSHMT2 was studied using glutaraldehyde cross-linking and non-denaturing PAGE. Shown are western blot images of cross-linked AtSHMT1 (A and B) and AtSHMT2 (C). Cross-linking time was 40 min (complete cross-linking) for A and 30min (partial cross-linking) for B and C. The cross-linked protein concentration (bottom of each lane) spanned the range used in the enzyme assays. The purified proteins wereseparated by SDS–PAGE on NuPAGE Novex� 4–12% Bis–Tris gels. The software myImage Analysis was used to measure the density of the dimer and tetramer bands on theexisting gels, and calculate the tetramer/dimer ratio from the ratio of the corresponding band densities. The tetramer/dimer ratio vs. enzyme concentration for cross-linkedAtSHMT1 (B) and AtSHMT2 (C) is shown below the PAGE gels. (D) Non-denaturing PAGE of AtSHMT1 (lane 1) and AtSHMT2 (lane 2). Sample load was 2 lg. NativeMARK™protein markers: 480-kDa apoferritin, 242-kDa B-phycoerythrin, and 146-kDa lactate dehydrogenase.

Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96 93

alone served as a positive control (Fig. 6A, D, and G). The discrepantresults of the proteomic studies [54–56] may have been caused bylimited purity of the analyzed plasma membrane preparations.

Discussion

SHMTs are the major entry points into the folate-mediated one-carbon metabolism in all organisms [14]. Functional SHMTs havebeen detected in mitochondria, plastids, nuclei, and the cytosolin plants [2,11,13,23], but only the plastid isozyme from A. thalianahas been biochemically characterized [10]. We here report func-

tional expression, purification, and biochemical characterizationof SHMTs from mitochondria (AtSHMT1 and AtSHMT2) and thecytosol (AtSHMT4) from A. thaliana.

While AtSHMT1 and AtSHMT4 appear to exist as single splicingvariants, two splicing variants of AtSHMT2 are present in Gen-Bank�; the longer variant (AtSHMT2long) has a 16-amino acidinsertion compared with the shorter one (AtSHMT2). Both variantswere expressed and purified following identical procedures, butonly AtSHMT2 was catalytically active. A previous study reportedthat expression of AtSHMT2long under the control of AtSHM1 pro-moter or CaMV 35S promoter cannot rescue the photorespiratoryphenotype of shm1 mutant [23]. Taken together, these results sug-

Fig. 5. Ribbon diagram of the modeled tetrameric AtSHMT2 and amino acidsequence alignment of AtSHMT1, AtSHMT2, and AtSHMT4. (A) One of the H4PteGlun

entry sites is zoomed up and the four regions (a–d) containing the major differencesof AtSHMT1 and AtSHMT2 to AtSHMT4 are highlighted in yellow. H4PteGlun andPLP are represented as ball and stick models. The inter-subunit salt bridges uniqueto both AtSHMT1 and AtSHMT2 are indicated in the ball and stick model. Thesefigures were generated using CCP4MG 2.5.0. (B) The amino acid residues consti-tuting the H4PteGlun entry site are marked with dotted boxes. The two residuesforming the inter-subunit salt bridge in SHMT1 and SHMT2 are marked with aclosed circle (d). The residues for formation of internal aldimine and for ion pairingwith PLP (Lys286 and Asp257) are marked with a closed triangle (N), the residuesinteracting with external aldimine (Tyr112, His260, and Arg430) with a closedsquare (j), the residues interacting with 50-phosphate of PLP (Tyr102 and His285)with an open circle (s), and the residues interacting with folate (Asn415 andTyr111) with an open triangle. The secondary structural elements are indicated inyellow (a-helix) and red (b-strand) on top of the corresponding sequences.

EGFP C-terminal N-terminal

Fig. 6. Transient expression of C- and N-terminal EGFP fusion to AtSHMT4. EGFPexpressed from the pUC18-GFP5T-sp plasmid (A, D, and G) served as a positivecontrol for targeting to the cytosol. The full-length AtSHMT4 was expressed as a C-terminal (B, E, and H) or an N-terminal (C, F, and I) fusion to EGFP. (A–C) EGFPfluorescence; (D–F) chlorophyll autofluorescence; (G–I) merged images.

94 Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96

gest that AtSHMT2long is probably catalytically inactive in vivo.Further study is needed to determine whether this variant has analternative physiological role.

Kinetic characterization showed that AtSHMT4 has the highestcatalytic efficiency (kcat/Km) of the three enzymes studied. Unlikethe mitochondrial isozymes, AtSHMT4 was also more susceptibleto substrate inhibition with the pentaglutamylated folate sub-strates, which are the most abundant folate derivatives in plantcells. Such differences in catalytic efficiency and sensitivity to sub-strate inhibition of the AtSHMTs may be interpreted as the adap-tion to folate concentrations in mitochondria and the cytosol. Aprevious study estimated that mitochondria and the cytosol plusnuclei contain respectively �50% and 40–45% of the total folatepool in pea leaves [7]. The volume occupied by mitochondria andthe cytosol has respectively been estimated at 2.5–4% and 17–25% of the total cell volume in spinach and barley leaves [57,58].Assuming these results apply to A. thaliana, we estimated its totalfolate pool to be at least 7.6-fold higher in mitochondria than in thecytosol. Thus, the higher catalytic efficiency of AtSHMT4 might bean adaption to the relatively low folate concentration in thecytosol.

Published values for folate concentration in pea leaf mitochon-dria are 0.4–1 mM [7,59]. Assuming further this result applies to A.thaliana, we estimated the total folate concentration at �700 lM inmitochondria and �90 lM in the cytosol. (S)-H4PteGlun and (S)-5,10-CH2–H4PteGlun together account for <5% of the total folatepool in any subcellular compartment in A. thaliana [25,60]. There-fore, concentration of (S)-H4PteGlun plus (S)-5,10-CH2–H4PteGlun

is expected to be �35 lM in mitochondria and �4.5 lM in thecytosol in A. thaliana. These estimates are higher than the corre-sponding Ki values for the three enzymes assayed with the penta-glutamylated folate substrates (Table 1) Therefore, substrateinhibition may affect the catalytic activity of these SHMTs in vivo.

Specific activity of AtSHMT1 and AtSHMT2 as a function of en-zyme concentration increased only when using monoglutamylatedfolate substrates. Previous studies have shown that SHMTs exist as

Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96 95

homodimers in prokaryotes [61–63] and as homotetramers ineukaryotes [41,64]. Two exceptions were a Bacillus stearothermo-philus SHMT [51] and the Asp89Asn mutant of a sheep liver SHMT[53], both existing as a mixture of dimers and tetramers, with theoligomerization state affecting catalytic activity. In the secondcase, the proportion of tetramers increased when PLP concentra-tion increased during purification, and the specific activity of thetetramer was over 20-fold higher than that of the dimer. A similarchange in specific activity was observed in phosphofructokinasesfrom mammals and yeast, which were more active when concen-trated [65,66]. The effect in phosphofructokinases is caused by en-zyme aggregation at higher concentrations, which stabilizes themore catalytically active conformation of the enzymes [66,67].

The presented results do not support the hypothesis that the in-creased specific activity of AtSHMT1 and AtSHMT2 at higher en-zyme concentration in the presence of monoglutamylated folatesubstrates is due to increased oligomerization because the ob-served oligomerization states did not change over the concentra-tion range studied (Fig. 4). One explanation for not detecting aconcentration-dependent change in oligomerization is that the re-sults from glutaraldehyde cross-linking underestimated the di-meric form of the enzymes by pulling dimers into tetramers. Totest this possibility, we carried out partial cross-linking experi-ments over a shorter incubation time (Fig. 4B and C). Analysis ofthese experiments suggested no conversion between dimeric andtetrameric forms of AtSHMT1 and AtSHMT2.

We considered analyzing the oligomerization state of these en-zymes using other independent methods, but found them unsuit-able. Gel filtration chromatography could not be used becausethis method relies on the protein of interest forming a ‘‘peak’’ dur-ing the analysis, which is incompatible with the need to control theenzyme concentration precisely. Analytical ultracentrifugation, dy-namic light scattering and fluorescence polarization were unsuit-able because they lacked the sensitivity needed to studyoligomerization at very low protein concentrations. Thus, glutaral-dehyde cross-linking was the only method we were able to useconsidering the need to keep the enzyme concentration constantand very low.

The mechanism responsible for the observed increase in specificactivity of AtSHMT1 and AtSHMT2 remains to be determined. Inthe absence of evidence for a change in the oligomerization state,we speculate that, in the presence of monoglutamylated folate sub-strates, these two enzymes exist in a balance of two tetramericconformational states differing in turnover rates, with the more ac-tive form being more abundant at higher enzyme concentrations.The existence of enzyme molecules having distinct activity statesdiffering in turnover rates has been described [68]. One may envi-sion a possible scenario where AtSHMT1&2 dimers associate toform two kinds of tetramers: one with lower activity and the otherwith higher activity. One may further envision the tetramer withthe lower activity having also a lower dissociation constant thanthe tetramer with the higher activity. Under these conditions, ascenario is possible where nearly all of the dimers exist withinlow-activity tetramers at low enzyme concentrations. As enzymeconcentration increases, the high-activity tetramers become moreabundant, thus causing an increase in the observed specific activityof the mixture.

The observed effect of enzyme concentration on specific activitywas absent in the presence of pentaglutamylated folate substrates.Thus, we hypothesize that the polyglutamyl tail binding stabilizesthe tetramers of AtSHMT1 and AtSHMT2 into a single activity state.Investigating the physiological significance of this effect in mito-chondria is beyond the scope of the present paper.

Structural modeling and sequence comparison (Fig. 5) suggestthat the constituting regions of the entry site for the folate sub-strates might have less rigidity and polarity in AtSHMT4. Such dif-

ferences may ease access to the catalytic site in the cytosolicenzyme compared to the mitochondrial ones, which is consistentwith the higher catalytic efficiency (kcat/Km) of AtSHMT4. Those re-gions at the folate entry site may also mediate the dependence ofactivity on enzyme concentration in AtSHMT1 and AtSHMT2 inthe presence of monoglutamylated folate substrates. Binding ofpolyglutamylated folate substrates might weaken the salt bridgespresent at the folate entry site in AtSHMT1 and AtSHMT2(Fig. 5A). In addition, the polyglutamate tail is expected to interactwith the polar side chains at the folate entry site. Both modes ofinteraction of the enzyme with the polyglutamate tail of folate sub-strates could induce conformational changes responsible for stabi-lizing the enzymes into a single activity state and eliminating theeffect of enzyme concentration on the activities of AtSHMT1 andAtSHMT2.

Conclusions

The biochemical characterization of AtSHMT1, AtSHMT2, andAtSHMT4 with respect to the impact of folate polyglutamylationon substrate saturation kinetics showed increased turnover ratesat higher enzyme concentrations in the presence of monoglutamy-lated folate substrates, but not in the presence of pentaglutamylat-ed folate substrates, for the two mitochondrial AtSHMTs, but notfor the cytosolic one. The oligomerization state did not change overthe range of enzyme concentration studied, suggesting that an-other mechanism is responsible for the observed change in activity.Modeling of the enzyme structures presented features that mayexplain the activity differences in AtSHMT1, AtSHMT2, andAtSHMT4.

Acknowledgments

This work was supported by National Science Foundation grantsMCB-0429968 and MCB-1052492 (to S.R.).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.abb.2013.06.004.

References

[1] L. Schirch, Adv. Enzymol. Relat. Areas Mol. Biol. 53 (1982) 83–112.[2] V. Besson, M. Neuburger, F. Rébeillé, R. Douce, Plant Physiol. Biochem. 33

(1995) 665–673.[3] A.D. Hanson, D.A. Gage, Y. Shachar-Hill, Trends Plant Sci. 5 (2000) 206–213.[4] V. Schirch, D.M. Szebenyi, Curr. Opin. Chem. Biol. 9 (2005) 482–487.[5] V. Schirch, S. Hopkins, E. Villar, S. Angelaccio, J. Bacteriol. 163 (1985) 1–7.[6] P. Gardeström, G. Edwards, D. Henricson, I. Ericson, Physiol. Plant. 64 (1985)

29–33.[7] M. Neuburger, F. Rébeillé, A. Jourdain, S. Nakamura, R. Douce, J. Biol. Chem. 271

(1996) 9466–9472.[8] S.R. Turner, R. Ireland, C. Morgan, S. Rawsthorne, J. Biol. Chem. 267 (1992)

13528–13534.[9] F. Rébeillé, M. Neuburger, R. Douce, Biochem. J. 302 (1994) 223–228.

[10] Y. Zhang, K. Sun, F.J. Sandoval, K. Santiago, S. Roje, Biochem. J. 430 (2010) 97–105.

[11] C.R. McClung, M. Hsu, J.E. Painter, J.M. Gagne, S.D. Karlsberg, P.A. Salomé, PlantPhysiol. 123 (2000) 381–392.

[12] A. Jamai, P.A. Salomé, S.H. Schilling, A.P. Weber, C.R. McClung, Plant Cell 21(2009) 595–606.

[13] N. Engel, R. Ewald, K.J. Gupta, R. Zrenner, M. Hagemann, H. Bauwe, PlantPhysiol. 157 (2011) 1711–1720.

[14] K.E. Christensen, R.E. MacKenzie, BioEssays 28 (2006) 595–605.[15] M. Luo, R. Orsi, E. Patrucco, S. Pancaldi, R. Cella, Plant Mol. Biol. 33 (1997) 709–

722.[16] J.M. Mouillon, S. Aubert, J. Bourguignon, E. Gout, R. Douce, F. Rébeillé, Plant J.

20 (1999) 197–205.[17] E.A. Cossins, L. Chen, Phytochemistry 45 (1997) 437–452.[18] A.D. Hanson, S. Roje, Rev. Plant Physiol. Plant Mol. Biol. 52 (2001) 119–137.

96 Z. Wei et al. / Archives of Biochemistry and Biophysics 536 (2013) 87–96

[19] V. Prabhu, K.B. Chatson, G.D. Abrams, J. King, Plant Physiol. 112 (1996) 207–216.

[20] R. Douce, M. Neuburger, Curr. Opin. Plant Biol. 2 (1999) 214–222.[21] E.A. Cossins, The Biochemistry of Plants, in: D.D. Davis (Ed.), Academic Press,

New York, 1987, pp. 317–353.[22] C.R. Somerville, W.L. Ogren, Plant Physiol. 67 (1981) 666–671.[23] L. Voll, A. Jamai, P. Renné, H. Voll, R. McClung, A. Weber, Plant Physiol. 140

(2006) 59–66.[24] V. Besson, F. Rébeillé, M. Neuburger, R. Douce, E.A. Cossins, Biochem. J. 292

(1993) 425–430.[25] P. Mehrshahi, S. Gonzalez-Jorge, T.A. Akhtar, J.L. Ward, A. Santoyo-Castelazo,

S.E. Marcus, A. Lara-Núñez, S. Ravanel, N.D. Hawkins, M.H. Beale, D.A. Barrett,J.P. Knox, J.F. Gregory 3rd, A.D. Hanson, M.J. Bennett, D. Dellapenna, Plant J. 64(2010) 267–279.

[26] V. Schirch, W.B. Strong, Arch. Biochem. Biophys. 269 (1989) 371–380.[27] E.A. Cossins, Folates and Pterins, in: R.L. Blakley, S.J. Benkovic (Eds.), John

Wiley & Sons, New York, 1984, pp. 1–59.[28] L. Schirch, Folates and Pterins, in: R.L. Blakley, S.J. Benkovic (Eds.), John Wiley

& Sons, New York, 1984, pp. 339–431.[29] W.B. Strong, S.J. Tendler, R.L. Seither, I.D. Goldman, V. Schirch, J. Biol. Chem.

265 (1990) 12149–12155.[30] C. Eckermann, J. Eichel, J. Schröder, Biol. Chem. 381 (2000) 695–703.[31] S. Ravanel, M.A. Block, P. Rippert, S. Jabrin, G. Curien, F. Rébeillé, R. Douce, J.

Biol. Chem. 279 (2004) 22548–22557.[32] F. Rébeillé, P. Gans, Plant Physiol. 88 (1988) 973–975.[33] H. Bauwe, U. Kolukisaoglu, J. Exp. Bot. 54 (2003) 1523–1535.[34] C.L. Ho, M. Noji, K. Saito, J. Biol. Chem. 274 (1999) 11007–11012.[36] Y. Zhang, K. Sun, S. Roje, Anal. Biochem. 375 (2008) 367–369.[37] R. Higuchi, B. Krummel, R.K. Saiki, Nucleic Acids Res. 16 (1988) 7351–7367.[38] J. Newman, Acta Crystallogr. D Biol. Crystallogr. 60 (2004) 610–612.[40] V.E. Fadouloglou, M. Kokkinidis, N.M. Glykos, Anal. Biochem. 373 (2008) 404–

406.[41] S.B. Renwick, K. Snell, U. Baumann, Structure 6 (1998) 1105–1116.[42] P. Emsley, B. Lohkamp, W.G. Scott, K. Cowtan, Acta Crystallogr. D Biol.

Crystallogr. 66 (2010) 486–501.[43] A.T. Brünger, P.D. Adams, G.M. Clore, W.L. DeLano, P. Gros, R.W. Grosse-

Kunstleve, J.S. Jiang, J. Kuszewski, M. Nilges, N.S. Pannu, R.J. Read, L.M. Rice, T.Simonson, G.L. Warren, Acta Crystallogr. D Biol. Crystallogr. 54 (1998) 905–921.

[44] J. Sheen, Plant Physiol. 127 (2001) 1466–1475.

[45] J.N. Scarsdale, G. Kazanina, S. Radaev, V. Schirch, H.T. Wright, Biochemistry 38(1999) 8347–8358.

[46] J.R. Jagath, B. Sharma, B. Bhaskar, A. Datta, N.A. Rao, H.S. Savithri, Eur. J.Biochem. 247 (1997) 372–379.

[47] D.M. Szebenyi, X. Liu, I.A. Kriksunov, P.J. Stover, D.J. Thiel, Biochemistry 39(2000) 13313–13323.

[48] W.N. Chang, J.N. Tsai, B.H. Chen, T.F. Fu, Protein Expr. Purif. 46 (2006) 212–220.[49] A.N. Bhatt, M.Y. Khan, V. Bhakuni, Protein Sci. 13 (2004) 2184–2195.[50] M.L. Urbanowski, M.D. Plamann, L.T. Stauffer, G.V. Stauffer, Gene 27 (1984)

47–54.[51] V.R. Jala, V. Prakash, N.A. Rao, H.S. Savithri, J. Biosci. 27 (2002) 233–242.[52] K.A. Zanetti, P.J. Stover, J. Biol. Chem. 278 (2003) 10142–10149.[53] J.V. Krishna Rao, J.R. Jagath, B. Sharma, N. Appaji Rao, H.S. Savithri, Biochem. J.

343 (1999) 257–263.[54] J. Ito, T.S. Batth, C.J. Petzold, A.M. Redding-Johanson, A. Mukhopadhyay, R.

Verboom, E.H. Meyer, A.H. Millar, J.L. Heazlewood, J. Proteome Res. 10 (2011)1571–1582.

[55] S.K. Mitra, B.T. Walters, S.D. Clouse, M.B. Goshe, J. Proteome Res. 8 (2009)2752–2767.

[56] A. Marmagne, M. Ferro, T. Meinnel, C. Bruley, L. Kuhn, J. Garin, H. Barbier-Brygoo, G. Ephritikhine, Mol. Cell. Proteomics 6 (2007) 1980–1996.

[57] H. Winter, D.G. Robinson, H.W. Heldt, Planta 193 (1994) 530–535.[58] H. Winter, D.G. Robinson, H.W. Heldt, Planta 191 (1993) 180–190.[59] B. Gambonnet, S. Jabrin, S. Ravanel, M. Karan, R. Douce, F. Rébeillé, J. Sci. Food

Agric. 81 (2001) 835–841.[60] E. Collakova, A. Goyer, V. Naponelli, I. Krassovskaya, J.F. Gregory 3rd, A.D.

Hanson, Y. Shachar-Hill, Plant Cell 20 (2008) 1818–1832.[61] S. Angelaccio, R. Chiaraluce, V. Consalvi, B. Buchenau, L. Giangiacomo, F. Bossa,

R. Contestabile, J. Biol. Chem. 278 (2003) 41789–41797.[62] S. Bhavani, V. Trivedi, V.R. Jala, H.S. Subramanya, P. Kaul, V. Prakash, N. Appaji

Rao, H.S. Savithri, Biochemistry 44 (2005) 6929–6937.[63] J.N. Scarsdale, S. Radaev, G. Kazanina, V. Schirch, H.T. Wright, J. Mol. Biol. 296

(2000) 155–168.[64] J.R. Jagath, B. Sharma, N.A. Rao, H.S. Savithri, J. Biol. Chem. 272 (1997) 24355–

24362.[65] J.J. Aragón, A. Sols, FASEB J. 5 (1991) 2945–2950.[66] J. Bär, O.H. Martínez-Costa, J.J. Aragón, Biochem. Biophys. Res. Commun. 167

(1990) 1214–1220.[67] G.D. Reinhart, H.A. Lardy, Biochemistry 19 (1980) 1484–1490.[68] D.M. Rissin, H.H. Gorris, D.R. Walt, J. Am. Chem. Soc. 130 (2008) 5349–5353.