The Arabidopsis thaliana fructokinase-like proteins FLN1 and FLN2 are required for the differentiation of plastids into photosynthetically competent chloroplasts. However, their specific roles are unknown. FLN1 and FLN2 localize in a multisubunit prokaryotic-type polymerase (plastid-encoded RNA polymerase) complex that transcribes genes encoding components of photosynthesis-related assemblies. Despite sequence identity with fructokinases, which are members of the pfkB (phosphofructokinase B) family of enzymes, kinase activity of FLN1 and FLN2 has not been demonstrated. Homology modeling using pfkB X-ray structures, sequence comparisons, and mutational analyses suggests that FLN proteins may bind their substrates differently from other pfkB proteins. We provide evidence that purified recombinant FLN1 undergoes an ATP-mediated change in binding affinity with both itself and recombinant FLN2. The ATP-mediated change in the affinity of FLN1 for FLN2 is not affected by mutations in conserved active-site residues known to affect catalysis in active pfkB enzymes. In contrast, recombinant FLN2 hetero-oligomerizes independently of ATP concentration. At ATP concentrations that promote FLN1 homomeric interactions, the FLN1–FLN2 hetero-oligomer is the dominant form in vitro. We further present evidence that FLN1 associates with a large protein complex in chloroplasts independently of ATP. Given that ATP levels fluctuate between light–dark cycles in the 1–5 mM range, we propose that changes in FLN1 and FLN2 interactions are biologically meaningful.
A ligand-induced change in a protein's affinity to its substrates or interacting macromolecules is commonly used to regulate enzyme activity or to alter the action of transcriptional regulatory proteins. There are many examples of this latter phenomenon. A classic system is the lactose operon in Escherichia coli in which the LacI repressor protein binds tightly to DNA in the absence of lactose. DNA binding and hence repression are reduced when allolactose is bound to LacI . In the yeast Kluyveromyces lactis, galactose functions as a ligand, regulating the expression of genes required for its catabolism. The repressor Gal80p binds to and inhibits the activator Gal4p in the absence of galactose. When galactose and ATP are bound to the galactokinase Gal1p, Gal1p then binds to Gal80p and their interaction blocks the inhibitory function of Gal80p on Gal4p. Intriguingly, this regulatory mechanism can operate with a catalytically inactive version of Gal1p, indicating that Gal1p has a dual function as a kinase and independently as a ligand-binding transcriptional regulatory protein .
In plants, several thylakoid-localized proteins required for photosynthetic light reactions are encoded by the chloroplast genome . Messenger RNAs coding for the aforementioned photosynthetic proteins are transcribed preferentially by a multisubunit, prokaryotic-type plastid-encoded RNA polymerase (PEP) . The active DNA-bound form of PEP is part of a larger protein complex known as the plastid transcriptionally active chromosome (TAC), which includes roughly 30 subunits that are referred to as pTAC proteins . The Arabidopsis thaliana fructokinase-like proteins FLN1 and FLN2 have both been identified as TAC components in proteomic surveys. FLN1 has been identified as a pTAC in multiple proteomic analyses, and so is considered a core component of the TAC complex [5–7]. FLN2 was identified as a pTAC in only one study , suggesting that FLN2 association with the TAC complex may be either transient or peripheral. Additionally, work from our laboratory and others suggests that both FLN1 and FLN2 must be present for optimal TAC activity and normal seedling greening [8,9].
FLN1 and FLN2 are members of the phosphofructokinase B (pfkB) family of proteins, so named due to their shared sequence motifs with the minor phosphofructokinase in E. coli, Pfk-2, encoded by the pfkB gene . This family is also called the ribokinase (RK) family, referring to another early characterized member . Most pfkB proteins studied function as carbohydrate or purine kinases, although substrates for all members are not known. Using the E. coli RK structure 1RKD as a representative model, pfkB proteins have two main domains: a larger active site containing an αβα domain consisting of a β-sheet sandwiched between two groups of α-helices and a smaller protruding single β-sheet known as the lid domain . RK and many other pfkB proteins function as dimers and interactions between monomers occur mainly across the lid domains. The lid has one longer β-strand that bends to align with the sheet in the other monomer as in the thumb of shaking hands. In the RK active site, there are fewer direct hydrogen bonds between the protein and the nucleotide phosphate than those between the protein and the carbohydrate substrate. This difference may account for the ability of some pfkB enzymes to utilize multiple nucleotide triphosphates as phosphoryl donors while being highly specific for their carbohydrate or purine substrate [12–14].
There are two signature motifs that define the pfkB family of proteins: an N-terminally located di-glycine (GlyGly) motif and a C-terminally located G/AXGD motif (reviewed in ref. ). The GlyGly motif is located in a turn that makes up a hinge between the αβα and lid domains [12,15–17]. Mutation of the second glycine to aspartate in the GlyGly motif of the pfkB protein adenosine kinase from Leishmania donovani led to reduced substrate affinity and a near loss of enzymatic activity . The aspartate in the G/AXGD motif is implicated in catalysis. One reaction mechanism employed in kinase reactions is the abstraction of a hydroxyl proton by a nearby aspartate, activating that oxygen for nucleophilic attack on the γ-phosphate of ATP [12,18]. Consistent with this model, substitution of the aspartate to arginine in Pfk-2 greatly reduced catalysis without affecting substrate binding .
Here, we present evidence that Arabidopsis FLN1 and FLN2 change from a preferred FLN2 homo-oligomer to a FLN1–FLN2 hetero-oligomer in response to changes in ATP concentration in vitro. Homology modeling provides a rationale for why FLN proteins lack fructokinase activity despite their overall high identity with fructokinases. We also confirm the association of FLN1 with a large protein complex independently of ATP concentration. These data led us to suggest a model in which FLN oligomerization responds to light-induced changes in stromal ATP. FLN oligomerization provides a potential mechanistic link between the observed up-regulation of PEP activity and increasing light intensity during chloroplast development.
Arabidopsis seeds (Col-0 ecotype) were surface-sterilized and grown in liquid growth media [4.3 g/l Murashige and Skoog basal salts (Sigma–Aldrich), 1% sucrose, 2.5 mM MES, and 1× B vitamins (0.5 µg/ml nicotinic acid, 1.0 µg/ml thiamine·HCl, 0.5 µg/ml pyroxidine·HCl, 0.1 µg/ml myo-inositol), pH 5.7].
Sequence analysis and homology modeling
Protein sequences of Arabidopsis FLN1 (At3g54090) and FLN2 (At1g69200) were downloaded from TAIR (The Arabidopsis Information Resource) genome annotation (v10, July 2016) and used to query the genomes of Chlamydomonas reinhardtii and Physcomitrella patens in Phytozome . Both species returned a single ORF, Cre13.g561850.t1.1 and Pp3c13_6840V3.1, respectively, referred to as Chlamydomonas and moss FLN-like proteins. For the Physcomitrella gene, the current annotation does not include a start codon. Manual inspection of the upstream sequence identified an in-frame start codon. When the ORF begins with this methionine, a protein sequence with a predicted chloroplast transit peptide (cTP) is generated , suggestive of chloroplast localization. This version of the gene was then analyzed for a predicted cTP cleavage site, and the version without the predicted cTP was used for alignments. All FLN-like proteins contain predicted cTPs, which were removed prior to sequence alignment in MUSCLE . We also included a sequence from a bacterial fructokinase (FRK) from Xylella fastidiosa because its crystal structure has been determined (PDB, 3LKI) and was used as a template for FLN homology modeling (described below). Alignments were highlighted using BOXSHADE. Homology models of FLN1 and FLN2 sequences lacking their predicted cTPs were generated in SWISS-MODEL  using the FRK structure 3LKI as a template. Active-site residues are depicted from Vibrio cholerae FRK (PDB, 5EYN) and Salmonella enterica aminoimidazole riboside (AIRs) kinase (PDB, 1TZ6). Figures depicting homology models and active-site residues were generated using VMD and Tachyon [23,24].
RNA was extracted from 7-day-old Arabidopsis seedlings using the RNeasy Plant Mini Kit (Qiagen) and used to generate cDNA using the SuperScript III First-Strand Supermix (Invitrogen). For the production of proteins in bacteria, primers were chosen that removed the predicted cTP from FLN1 and FLN2 (site predicted using ChloroP, denoted as mFLN1 and mFLN2 for mature FLN1 and FLN2, respectively), while adding gateway (Invitrogen) recombination sites. Primers for mFLN1 were 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTGGAAGGAGATAGAACCATGGCTTCAATTAATGGCAGCG-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTC CTACCACATCGATGGAACATAAAC-3′. Primers for mFLN2 were 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTGGAAGGAGATAGAACCATGGCTGCTGGTAGGAGAAAG-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTC TCATAAGCTTCCATCTTCAAACATTG-3′. PCR products for all ORFs were recombined into pDONR201 (Invitrogen) via the BP clonase (Invitrogen) reaction. ORFs were then recombined into pEAK2 or pEAK4  for bacterial expression via the LR clonase (Invitrogen) reaction. Vectors containing FLAG-mFLN1 ORFs encoding proteins with single amino acid substitutions were generated from a FLN1 ORF harboring both substitutions synthesized by Genewiz (Callis lab clone p8535). FLAG-mFLN1G148D ORF was made via digestion of the wild-type FLN1 ORF in pDONR201 with HindIII/EcoRI and replacing that fragment with the equivalent HindIII/EcoRI fragment from p8535 encoding one substitution. FLAG-mFLN1D405R ORF was generated via digestion of the double-mutant FLN1 ORF in p8535 with HindIII/EcoRI and ligating in the equivalent fragment from the wild-type FLN1 ORF from pDONR201. Once both single-mutant mFLN1 cDNA pDONR plasmids were generated, the identical procedure as described above for the wild-type coding regions was used to create vectors for production of recombinant single-mutant FLAG-mFLN1s.
Preparation of recombinant proteins and GST pulldown assays
Proteins were produced in E. coli strain BL21-LEMO (New England Biolabs) and enriched either on Ni-sepharose (GE Healthcare) or loaded onto Glutathione Sepharose (GE Healthcare). For Ni-sepharose chromatography of pEAK2-based clones, bacterial cells were lysed by sonication in lysis buffer [20 mM phosphate, 500 mM NaCl, and 30 mM imidazole (pH 7.4)] washed with at least 300 column volumes of wash buffer [20 mM phosphate, 1.5 M NaCl, and 30 mM imidazole (pH 7.4)] and eluted in elution buffer [20 mM phosphate, 250 mM NaCl, and 500 mM imidazole (pH 7.4)]. Glycerol was added to a final concentration of 20% v/v and proteins were flash-frozen. For glutathione–sepharose chromatography of pEAK4-based clones, bacterial cells were lysed by sonication in lysis buffer [50 mM Tris and 500 mM NaCl (pH 7.5)], bound to beads, and the beads were washed with at least 10 column volumes of wash buffer [50 mM Tris and 300 mM NaCl (pH 7.5)] and the bead-bound protein was flash-frozen in storage buffer [50 mM phosphate, 150 mM NaCl, and 20% glycerol (pH 7.5)]. Recombinant protein concentrations were determined using Protein Assay Reagent (BioRad). Pulldown reactions were carried out in 300 µl reaction volume in PD buffer [20 mM Tris, 0.1 mM EDTA, 100 mM NaCl, 10 mM MgCl2, and 0.2% Triton X-100 (pH 7.5)]. Approximately 3 µg of GST fusion protein immobilized on glutathione beads was added to the reaction tube and mixed with ∼3 µg of soluble FLAG fusion protein. Loading control aliquots were removed, and then nucleotides were added to the indicated concentration. Reactions were incubated on a rotating mixer at 4°C for 1 h, washed in PD buffer + nucleotide (at the same concentration in the binding reaction), boiled in Laemmli sample buffer [200 mM Tris (pH 6.8), 32% (v/v) glycerol, 6.4% (w/v) SDS, 0.32% (w/v) Bromophenol Blue, and 200 mM DTT], separated on a 10% acrylamide SDS–PAGE gel, and then transferred to a PVDF membrane (Millipore). The membrane was blocked in Tris-buffered saline with 0.1% Tween-20 and 5% non-fat dried milk for 15 min at room temperature, and analyzed by Western blot with various antibodies. Antibodies and concentrations used in these experiments: anti-FLAG-HRP (1°, 1 : 2500, Sigma), anti-GST (1°, 1 : 5000, Santa Cruz Biotechnology), and goat-anti-rabbit-HRP (2°, 1 : 5000, Jackson ImmunoResearch). All blots were visualized using the ECL Western Blotting Detection Reagent (GE Healthcare).
Chloroplasts were prepared as described previously  with minor changes. Seven-day-old Arabidopsis seedlings were kept in the dark overnight prior to chloroplast isolation. Plant material was chopped using scissors in grinding buffer [0.05 M HEPES, 0.33 M sorbitol, 2 mM EDTA, and 1% w/v BSA (pH 8.0)] and then blended three times for 5 s each in a Waring blender. The lysate was then passed through two layers of Miracloth (Millipore) and centrifuged over a 50% Percoll (GE Healthcare) gradient at 8000×g for 10 min at 4°C. Isolated chloroplasts were resuspended to 0.5 mg/ml chlorophyll in assay buffer (0.05 M HEPES and 0.33 M sorbitol (pH 8.0)] and either ATP or glycerate was added to a final concentration of 10 mM; then the mixture was incubated for 30 min at 4°C in the dark. Chloroplasts were isolated and treated identically for the determination of the effect of glycerate on ATP levels, with the exception that no ATP was added to the control sample. Isolated chloroplasts were then lysed in a hypotonic buffer (0.05 M Bis–Tris, 0.5 M 6-aminocaproic acid, 10% glycerol, and 1% Triton-X 100), and Quantilum luciferase (Promega) was used to measure ATP concentration against a standard curve according to the manufacturer's instructions.
2D Blue native PAGE
Arabidopsis chloroplasts prepared as described above and treated with 10 mM ATP or 10 mM glycerate were lysed in a hypotonic buffer (0.05 M Bis–Tris, 0.5 M 6-aminocaproic acid, 10% glycerol, and 1% Triton-X 100) supplemented with either 10 mM ATP or 10 mM glycerate and centrifuged at 16 000×g at 4°C for 20 min to remove insoluble materials. A 50 µl aliquot of chloroplast lysate was loaded onto a 4–15% gradient native polyacrylamide gel (50 mM Bis–Tris and 0.5 M 6-aminocaproic acid) and run at 30 V overnight at 4°C using anode buffer [50 mM Bis–Tris (pH 7.0)] and blue cathode buffer [50 mM Tricine, 15 mM Bis–Tris, and 0.02% Brilliant blue G (Sigma–Aldrich), pH 7.0]. Gel lanes were excised, treated in reducing buffer [65 mM Tris–HCl, 3.3% SDS, and 4% 2-mercaptoethanol (pH 6.8)] at 50°C for 45 min and then washed in SDS–PAGE running buffer for 10 min. They were then placed horizontally on top of the stacking portion of an SDS gel and fractionated by standard SDS–PAGE. Proteins were transferred to PVDF membranes (Millipore) and Western blots were carried out as described for GST pulldown assays, but using affinity-purified polyclonal anti-FLN1 antibodies that were raised in rabbits against the mature recombinant FLN1 protein by Proteintech (http://www.ptglab.com/). Anti-FLN1 antibodies were used at a 1:5000 dilution and secondary goat-anti-rabbit antibodies conjugated to horseradish peroxidase were used at a 1:5000 dilution.
FLN1 and FLN2 sequences contain insertions when compared with an active fructokinase
While phylogenetic trees place FLN1 and FLN2 most closely with active fructokinases [8,14], we and others  have been unable to detect fructokinase activity from purified recombinant FLN1 and FLN2 in vitro. To begin to understand differences between FLNs and active enzymes, we compared the peptide sequences of Arabidopsis FLN1 and FLN2 (AtFLN1 and AtFLN2) and FLN-like proteins from moss (P. patens) and C. reinhardtii to an active FRK from the bacterial species X. fastidiosa, whose X-ray structure is known (Figure 1A). All eukaryotic protein sequences were predicted to contain an N-terminal cTP, which was removed for our analyses (referred to here as mFLN proteins, for mature). Moss and Chlamydomonas proteins were included because both genomes are predicted to encode a single FLN-like protein, which may represent an evolutionary precursor to higher plant FLNs. In contrast, all vascular plants investigated encode orthologs of both AtFLN1 and AtFLN2 .
FLN proteins have multiple sequence inserts when compared with active FRKs.
FLN proteins contain the motifs characteristic of pkfB proteins . The GlyGly motif is conserved in all except the moss FLN-like protein, which contains a threonine instead of the second glycine. The G/AXGD motif is present in all protein sequences in our analysis. However, based on an alignment (Figure 1A), there are three sites with additional amino acids, conserved among the mFLN and mFLN-like proteins, that are not present in bacterial FRKs. First, mFLNs have additional N-terminal amino acids, hereafter referred to as the N-terminal extension, which contain no significant identity with any known functional domain. AtFLN2 and the moss FLN-like protein have the longest N-terminal extensions, followed by AtFLN1. The Chlamydomonas FLN-like protein has the shortest N-terminal extension. The second site of amino acid insertion occurs between residues 27 and 28 in bacterial FRK (Figure 1A, underlined in blue). Again, the Chlamydomonas FLN insertion is the shortest, here with only a single additional amino acid. AtFLN1 and AtFLN2 harbor 15 additional amino acids, and the moss FLN-like protein has 18. The moss protein insertion is more similar in sequence to AtFLN1's insertion than to that of AtFLN2. The third amino acid insertion site in the FLN and FLN-like proteins relative to FRK occurs between bacterial FRK residues 206 and 214 (Figure 1A, underlined in orange). The most similar insertions are again those in moss and AtFLN1 with 34 and 33 additional amino acids, respectively, when compared with the bacterial FRK. AtFLN2 has 22 additional amino acids and Chlamydomonas again has the smallest insertion with only four amino acids.
To understand the possible consequences of these insertions, we utilized the SWISS-MODEL tool  to generate homology models of both AtFLN1 (Figure 1B, blue structure) and AtFLN2 (Figure 1C, blue structure) using the known structure of the X. fastidiosa FRK (3LKI) as a template (Figure 1B,C, red structure). Neither the AtFLN1- and AtFLN2-predicted cTP nor their N-terminal extensions were included in the homology model as FRK lacks a homologous sequence. The modeling indicates that FLN proteins are predicted to contain all the structural elements of FRK, with the two internal insertions clearly accommodated in loops between FRK secondary structure elements in the models when compared with the bacterial FRK structure. In both AtFLN1 and AtFLN2, the first internal insertion (Figure 1A, underlined in blue) is positioned between two β-strands in the lid domain (Figure 1B,C, circled in blue). In both AtFLN1 and AtFLN2, the second, larger, internal insertion is between two α-helices in the αβα domain and is predicted to point away from the lid domain (Figure 1B,C, circled in orange). Since the lid domain is the site of dimerization, the insertion in the lid domain is of particular interest as it may alter the way the FLN proteins oligomerize.
Amino acid substitutions in the AtFLN1 active site may alter substrate binding
To gain further insights into how FLN proteins might bind their substrates, we examined the active-site residues of two other pfkB enzymes that have had their structures determined with substrates in their respective active sites. We focused on comparing AtFLN1 sequences with these structures rather than FLN2 as FLN1 loss results in a more severe plant growth phenotype [8,9]. The pfkB proteins V. cholerae FRK (PDB 5EYN, unpublished) and S. enterica AIR kinase (PDB 1TZ6)  were chosen because both structures were determined with ATP analogs and fructose or AIRs in their active sites, respectively. Residues close to, and probably making contact with, the substrate molecules in both structures were identified via a distance search. Eight residues were identified in the search, with two probably contacting the nucleotide and six probably contacting the sugar (Figure 2). Residues in AtFLN1 corresponding to those identified in the pfkB structures were identified via sequence alignment.
PfkB active-site residues are not conserved in FLN1.
One of the two residues contacting the nucleotide, Thr234 (5EYN), is conserved in AtFLN1 and corresponds to AtFLN1 Thr368 (Figure 2A, blue underlined and C, bottom row, left). The corresponding residue in 1TZ6 is Ser220 (Figure 2B), suggesting that the hydroxyl group in this position is important. The other residue is an asparagine in both structures (296 in 5EYN and 281 in 1TZ6) that corresponds to AtFLN1 Val436 (Figure 2A,B, in blue, parentheses denote corresponding residue in AtFLN1, and C, bottom row, right). Replacement of the polar asparagine with a slightly smaller, aliphatic valine in AtFLN1 would result in a loss of hydrogen-bonding contacts and would probably alter the way ATP is bound at this site in AtFLN1.
AtFLN1 Asp405 and the corresponding aspartate residues in the pfkB structures are part of the G/AXGD motif important for catalysis (Figure 2A,B, red underlined, and C, bottom row, middle alignment). This residue is conserved in AtFLN1, AtFLN2, and in all plant FLN-like proteins evaluated (Figure 1A) . Of the other residues contacting the sugar substrate in the active enzymes, only two are found in FLN1. The asparagine in the prokaryotic structures at 49 in 5EYN and 35 in 1TZ6 (Figure 2A,B, below active site, black underlined) is Asn153 in AtFLN1 (Figure 2C, top row, middle alignment). Another asparagine, residue 174 in 5EYN and 160 in 1TZ6 (Figure 2A,B, above active site, black underlined), is conserved in AtFLN1 at Asn280 (Figure 2C, top row, right).
Three substrate-interacting residues are not conserved in AtFLN1. These are two aspartate residues (26 in 5EYN, 12 in 1TZ6 and 30 in 5EYN, and 16 in 1TZ6) and an arginine residue (176 in 5EYN and 162 in 1TZ6). The corresponding residues in AtFLN1 are Ala109, Glu113, and Pro282, respectively (Figure 2C, top row, left alignment for the first two, top right for Pro282). The replacement of aspartate with alanine in AtFLN1 would result in a loss of hydrogen-bonding contacts and could create a slightly larger space in the active site between the protein and a possible sugar substrate, compared with active pfkB enzymes. Aspartate replaced with glutamate retains the polarity of the residue at that position, and the larger side chain of glutamate may be compensatory for the nearby, aforementioned, aspartate-to-alanine replacement. The substitution of proline in AtFLN1 for arginine in the active enzymes would again result in fewer hydrogen-bonding contacts in AtFLN1 and may actually alter the protein structure due to the steric limitations on proline residues.
Interactions between AtFLN proteins are modulated by ATP
To determine whether the additional residues near the lid region in FLN proteins affect protein–protein interactions, we produced recombinant GST- and FLAG-tagged AtFLN1 and AtFLN2 lacking their predicted cTPs  (denoted as mFLN1 and mFLN2 for mature FLN1 and FLN2, respectively). These proteins were used to study the in vitro interactions between AtFLN proteins using GST pulldown assays. The ability of the FLN proteins to form homo-oligomers was studied first. FLAG-mFLN1 interacted slightly with bound GST-mFLN1 without added ATP and interacted at this same low level in concentrations of ATP at or below 3 mM (Figure 3A). At 4 and 5 mM ATP, the interaction between these two proteins was stronger (Figure 3A), suggesting that mFLN1 may interact with ATP and that interaction increases the affinity between mFLN1 molecules. In contrast, FLN2 self-association does not appear to be affected by ATP as FLAG-mFLN2 interacted with GST-mFLN2 at similar levels without ATP and in all concentrations of ATP tested (Figure 3B).
Interactions between mFLN proteins respond to ATP concentration.
Hetero-oligomerization between FLN1 and FLN2 was assessed next. FLAG-mFLN1 and GST-mFLN2 interaction was strongly stimulated in concentrations of ATP above 3 mM (Figure 3C). Because both FLN proteins have been detected in proteomic studies [5–7], indicating both are present in vivo, we assayed for their interaction in a competition test, comparing the ability of FLAG-mFLN1 to displace FLAG-FLN2 bound to GST-mFLN2 at different ATP concentrations. As their mobilities are distinct by SDS–PAGE, a single antibody can detect both FLN proteins tagged with the same epitope. The competition assay was performed in a two-step process. We first incubated bead-bound GST-mFLN2 in the presence of FLAG-mFLN2 to create the ATP-independent mFLN2 homo-oligomer. We then added equivalent amounts of these beads containing the mFLN2 homo-oligomers to several test tubes and added soluble FLAG-mFLN1 in various ATP concentrations. The mFLN1/mFLN2 ratios were equivalent across all conditions. An interaction between GST-mFLN2 and FLAG-mFLN2 was present without ATP and at low ATP concentrations, but as the concentration of ATP increased to 4 mM and above, the amount of FLAG-mFLN1 recovered by interaction with GST-mFLN2 increased, with a concomitant reduction in the recovery of FLAG-mFLN2 (Figure 3D). These data indicate that FLN2 probably exists as a homo-oligomer when ATP levels are low, but as ATP levels increase, there is a switch in affinity that favors the hetero-oligomerization of FLN1 and FLN2. We verified that both FLAG-tagged proteins did not bind bead-bound GST alone in any ATP concentration (Figure 3E,F).
Interactions between recombinant mFLN1 and mFLN2 are specific for the nucleotide ATP
The ATP/ADP ratio regulates photosynthetic activity in chloroplasts , and GTP regulates protein translocation across chloroplast membranes . Chloroplasts also contain other nucleotides for incorporation into chloroplast RNA and DNA. There is also evidence that related proteins, Arabidopsis fructokinases, are able to utilize other nucleoside triphosphates in catalysis, albeit with lower efficiencies, suggesting they bind to other nucleotides . Because these other nucleosides and nucleotides are present in vivo, we determined whether some of these compounds stimulated hetero-oligomerization between FLN proteins (Figure 4). In each case, the same protein preparations were assayed for interaction in increasing ATP in positive control parallel experiments. Interactions between FLAG-mFLN1 and GST-mFLN2 in 1 or 3 mM ADP were not detectable, equivalent to 1 or 3 mM ATP-containing assays performed at the same time. In contrast with samples at 5 mM ATP, interaction between FLAG-mFLN1 and GST-mFLN2 was not detectable in the presence of 5 mM ADP (Figure 4A). While this result does not indicate whether FLN hetero-oligomerization is responsive to the ATP/ADP ratio as are other processes in chloroplasts, it indicates that ADP alone, at levels in excess of its physiological maximum, does not stabilize the interaction between mFLN1 and mFLN2 to nearly the same extent as ATP does in vitro. The interaction between FLAG-mFLN1 and GST-mFLN2 was undetectable in 5 mM GTP (Figure 4B). Similar results were obtained when using AMP, CTP, and UTP at identical concentrations (Figure 4C–E, respectively), while parallel ATP experiments showed increased interaction at 5 mM ATP, as expected.
Interaction between FLN1 and FLN2 is ATP-specific.
FLN-mediated ATP hydrolysis has not been detected using recombinant protein in in vitro assays , so we addressed the question of whether ATP hydrolysis is required for the stimulation of interaction by using ATP analogs that are not hydrolyzable or slowly hydrolyzed. Neither AMP-PNP nor ATPγS stimulated the interaction between FLAG-mFLN1 and GST-mFLN2 at concentrations up to 5 mM (Figure 4F,G, respectively). The finding that non-hydrolyzable and slowly hydrolyzable ATP analogs do not stimulate interactions between recombinant FLN proteins in vitro suggests either that ATP hydrolysis is required for the interaction between proteins to take place, or that the structures of the analogs, though very similar to ATP, either do not bind to the protein or bind in such a way that does not support oligomerization of the FLN proteins. To differentiate between these two models, we generated recombinant FLAG-mFLN1 carrying either a G148D or D405R mutation, referred to as FLAG-mFLN1G148D and FLAG-mFLN1D405R, respectively, and tested whether they interact with GST-mFLN2 in an ATP-dependent manner in GST pulldown assays. Mutations in either site have been shown to reduce the activity of FRKs below the limits of detection [15,18]. A similar mutation, D405N, in particular, did not appreciably change ligand binding, though the enzyme was inactive . Both FLAG-mFLN1G148D and FLAG-mFLN1D405R interacted with GST-mFLN2 in an ATP-dependent manner (Figure 5A,B, respectively). They also showed nearly identical ATP concentration dependence on their interaction with GST-mFLN2 as the wild-type protein (Figure 5). From these data, we conclude that ATP hydrolysis is not required for the interaction between recombinant AtFLN proteins in vitro.
Interaction between putative active-site FLN1 mutants responds to ATP concentration.
FLN1 interacts with a large protein complex in Arabidopsis chloroplasts
Data from multiple studies of the TAC complex show members thought to be associated with the intact complex migrating in blue native (BN) gels at molecular mass >900 kDa [7,30]. A proteomic analysis designed to identify proteins tightly associated with the core subunits identified FLN1 . We tested whether this association of FLN1 with its protein complex is responsive to ATP concentrations. We isolated chloroplasts from 7-day-old, liquid grown, wild-type Arabidopsis seedlings, and either supplemented the media with ATP or depleted the ATP chemically via the addition of glycerate . We verified that our treatment depleted ATP via luciferase assays in lysates from chloroplasts treated with and without 10 mM glycerate (Figure 6A). Chloroplast protein complexes from lysates were fractionated by two-dimensional BN PAGE. Native protein complexes were separated in the first dimension, and then they were separated into their constituents in the SDS-containing second dimension . Using an antibody raised against recombinant Arabidopsis mFLN1, we observed that endogenous FLN1 migrated in a >900 kDa complex in chloroplasts treated with 10 mM ATP (Figure 6B). In chloroplasts treated with 10 mM glycerate, which reduced ATP levels, the size of the complex that FLN1 was associated with did not change appreciably, and no LMW complexes containing FLN1 were detected (Figure 6C).
FLN1 migrates as part of a large protein complex in 2D PAGE regardless of ATP concentration.
FLN proteins are conserved in the plant lineage, from Chlamydomonas to angiosperms. They are known components of the chloroplast PEP complex and are required for PEP-dependent transcription. Loss-of-function fln mutants are either slow greening or fail to green [6,8,9,33,34], indicating their significance in plant biology. Despite this significance, the molecular mechanism of FLN function is unknown. Both AtFLN proteins interact with a novel plastid-localized thioredoxin, TRXZ . TRXZ has thioredoxin activity in vitro and inactivation of TRXZ in Arabidopsis is seedling-lethal. As AtFLNs' interactions with TRXZ are dependent on specific FLN cysteine residues, a hypothesis was proposed that FLN proteins serve as redox sensors in vivo. However, the putative redox active residues on both TRXZ and AtFLN1 that mediate their interactions were previously reported to be dispensable in vivo under normal growth conditions , rendering this hypothesis less appealing.
FLN proteins also interact with other TAC components, but the biochemical consequences of these interactions are unexplored. AtFLN1 binds in vitro to pTAC7, another essential TAC component that, in turn, interacts with several other subunits . AtFLN2 binds to pTAC5 , which interacts with HSP21. Both AtFLN2 and pTAC5 appear to be essential for development under heat stress . Since the functions of these FLN interactors are uncharacterized, the effect of FLN binding on their activities cannot be tested.
FLN proteins belong to a family of active carbohydrate and purine kinases, which suggests that their catalytic activity may be critical for function. Despite the highest identity with fructokinases from among the pfkB proteins, no corresponding activity has been detected . Molecular modeling predicts that FLNs contain all the secondary elements in a known fructokinase and that they can be threaded on an active fructokinase template. However, FLN proteins contain additional N-terminal and internal amino acids accommodated in loops that may interfere with canonical pfkB-type interactions and substrate binding. It is tempting to speculate that these additional regions may provide contacts for interactions with pTACs mentioned above or may serve some other regulatory role.
Molecular modeling also predicts that FLN proteins lack some residues involved in substrate interactions in active pfkB enzymes. The residues around the sugar molecule in pfkB structures, in particular, are very different. The replacement of aspartate and arginine with alanine and proline, respectively, in AtFLN1 results in a loss of several hydrogen bonds between the protein and the sugar molecule. Glu113 in AtFLN1, which corresponds to aspartate residues in the active enzymes, is a conservative mutation as it retains the polarity and charge of the original residue, though the larger side chain probably extends slightly further into the active site and would result in a change in the landscape therein, which, in turn, would be predicted to result in altered substrate-binding properties of the protein. This poor conservation between AtFLN1 residues in the vicinity of the fructose/AIR suggests either altered substrate binding or no substrate binding in AtFLN1 and, by extension, in FLN-like proteins generally. Though the catalytic Asp405 is present in AtFLN1 and all other FLN-like proteins examined herein, the lack of several residues that orient the substrates in such a way that supports catalysis provides a possible explanation for the lack of observed enzymatic activity.
Previously, pulldown experiments with recombinant protein detected FLN1–FLN2 interactions . Here, we extend those studies to demonstrate that FLNs exhibit both homo- and heteromeric protein–protein interactions and, significantly, that these interactions are affected by ATP. Our data indicate that, in 3–5 mM ATP, an FLN1–FLN2 hetero-oligomer is preferred, while an FLN2 homo-oligomer predominates in <3 mM ATP in vitro. ATP changes do not dramatically affect FLN1's association with a large complex. The lack of a difference in the ATP-dependent oligomerization of the FLN proteins harboring substitutions in the active site or in one involved in lid movement also supports the hypothesis that sugar binding is altered or absent from FLN-like proteins when compared with active pfkB enzymes. ATP is a product of photosynthesis and, as such, cycles in the light and dark, as does expression of PEP-dependent genes . However, the molecular mechanism for control of the latter is not well defined. Though the biological significance of FLN oligomerization is not yet known, our data lead to an attractive model in which FLN proteins, through their ATP-dependent oligomerization, may regulate the activity of the TAC complex and, therefore, transcription of plastidic genes related to photosynthesis. Further studies are required to support or reject many aspects of this proposed model, though our data provide a basis for those studies and bring to light an interesting biochemical property of the enigmatic FLN proteins.
chloroplast transit peptide
fructokinase (EC 188.8.131.52)
low molecular weight
open reading frame
plastid-encoded RNA polymerase
minor phosphofructokinase in E. coli
transcriptionally active chromosome
J.W.R. and J.C. designed the project and experiments. J.W.R. performed the experiments. J.W.R. and J.C. wrote the manuscript.
This work was funded by the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences of the U.S. Department of Energy through Grants [DE253FG02-09ER16077] and [DE-SC0002175].
The authors thank Kentaro Inoue, Li Tian, Clark Lagarias, and Nathan Rockwell for their helpful comments on the manuscript. We also thank Nathan Rockwell for preparing Figure 2. We thank Katrina Linden and Charles Gasser for careful reading of the manuscript.
The Authors declare that there are no competing interests associated with the manuscript.