Structure of the Cyanobacterial Magnesium Chelatase H Subunit Determined by Single Particle Reconstruction and Small-angle X-ray Scattering*

Background: In chlorophyll biosynthesis, ChlH is the protoporphyrin-binding subunit of magnesium chelatase. Results: ChlH has an ∼132-kDa domain with a cavity, connected to an ∼16-kDa domain. Conclusion: ChlH could enclose the magnesium protoporphyrin product, chaperoning it to the next enzyme in the pathway. Significance: The structure of ChlH will help to unravel its multiple catalytic and regulatory functions. The biosynthesis of chlorophyll, an essential cofactor for photosynthesis, requires the ATP-dependent insertion of Mg2+ into protoporphyrin IX catalyzed by the multisubunit enzyme magnesium chelatase. This enzyme complex consists of the I subunit, an ATPase that forms a complex with the D subunit, and an H subunit that binds both the protoporphyrin substrate and the magnesium protoporphyrin product. In this study we used electron microscopy and small-angle x-ray scattering to investigate the structure of the magnesium chelatase H subunit, ChlH, from the thermophilic cyanobacterium Thermosynechococcus elongatus. Single particle reconstruction of negatively stained apo-ChlH and Chl-porphyrin proteins was used to reconstitute three-dimensional structures to a resolution of ∼30 Å. ChlH is a large, 148-kDa protein of 1326 residues, forming a cage-like assembly comprising the majority of the structure, attached to a globular N-terminal domain of ∼16 kDa by a narrow linker region. This N-terminal domain is adjacent to a 5 nm-diameter opening in the structure that allows access to a cavity. Small-angle x-ray scattering analysis of ChlH, performed on soluble, catalytically active ChlH, verifies the presence of two domains and their relative sizes. Our results provide a basis for the multiple regulatory and catalytic functions of ChlH of oxygenic photosynthetic organisms and for a chaperoning function that sequesters the enzyme-bound magnesium protoporphyrin product prior to its delivery to the next enzyme in the chlorophyll biosynthetic pathway, magnesium protoporphyrin methyltransferase.

The biosynthesis of chlorophyll, an essential cofactor for photosynthesis, requires the ATP-dependent insertion of Mg 2؉ into protoporphyrin IX catalyzed by the multisubunit enzyme magnesium chelatase. This enzyme complex consists of the I subunit, an ATPase that forms a complex with the D subunit, and an H subunit that binds both the protoporphyrin substrate and the magnesium protoporphyrin product. In this study we used electron microscopy and small-angle x-ray scattering to investigate the structure of the magnesium chelatase H subunit, ChlH, from the thermophilic cyanobacterium Thermosynechococcus elongatus. Single particle reconstruction of negatively stained apo-ChlH and Chl-porphyrin proteins was used to reconstitute three-dimensional structures to a resolution of ϳ30 Å. ChlH is a large, 148-kDa protein of 1326 residues, forming a cage-like assembly comprising the majority of the structure, attached to a globular N-terminal domain of ϳ16 kDa by a narrow linker region. This N-terminal domain is adjacent to a 5 nm-diameter opening in the structure that allows access to a cavity. Smallangle x-ray scattering analysis of ChlH, performed on soluble, catalytically active ChlH, verifies the presence of two domains and their relative sizes. Our results provide a basis for the multiple regulatory and catalytic functions of ChlH of oxygenic photosynthetic organisms and for a chaperoning function that sequesters the enzyme-bound magnesium protoporphyrin product prior to its delivery to the next enzyme in the chlorophyll biosynthetic pathway, magnesium protoporphyrin methyltransferase.
Photosynthesis supports the vast majority of life on Earth and chlorophyll a (Chl) 2 is the essential cofactor for this process (1). Thus, it is important to understand the mechanisms and regulation of the biosynthetic pathway for Chls. Heme and Chl pigments share a common biosynthetic pathway that diverges at the point where either Fe 2ϩ or Mg 2ϩ , respectively, is inserted into protoporphyrin IX (Proto). Insertion of Fe 2ϩ is catalyzed by ferrochelatase (EC 4.99.1.1), whereas the three-subunit enzyme magnesium chelatase (E.C. 6.6.1.1, Mg chelatase) catalyzes the ATP-dependent formation of Mg-protoporphyrin IX (MgProto). This reaction was first demonstrated for the Mg chelatase from the purple photosynthetic bacterium Rhodobacter (R.) sphaeroides using recombinant bchI, D, and H genes overexpressed in Escherichia coli (2), and subsequently it was shown that Mg chelatase activity could be reconstituted using recombinant I, D, and H subunits from the cyanobacterium Synechocystis (3), the green sulfur bacterium Chlorobium vibrioforme (4), and from tobacco (5). In each case the I, D, and H subunits have molecular weights of 38 -42 kDa, 60 -74 kDa, and 140 -150 kDa, respectively. The I subunit is an ATPase and forms a complex with the D subunit (6 -9), whereas the H subunit binds both Proto and MgProto (10). Recently, the formation of a ChlH⅐ChlI⅐ChlD Mg chelatase complex has been demonstrated (11).
Steady-state assays of Mg chelation have demonstrated that this process shows a cooperative response to Mg 2ϩ concentration and is accompanied by the hydrolysis of 14 ATP molecules (12). The kinetic analysis of this enzyme complex was extended using a transient-state approach that showed that nucleotide hydrolysis occurs after the rate-determining step, suggesting a mechanism where nucleotide binding clamps the chelatase in a product complex. Furthermore, it was proposed that at high porphyrin concentrations the first turnover promotes the enzyme into a more active state (13).
Structural studies of Mg chelatase started with the determination of the structure of the R. capsulatus BchI subunit to 2.1 Å (14). The Mg chelatase I subunits can be grouped within the ATPases associated with various cellular activities (AAA ϩ ) family, which often forms multimeric complexes. Subsequently, single-particle reconstruction methods using electron microscopy (EM) of negatively stained BchI complexes demonstrated that the I subunit can form hexameric rings (15) and heptameric rings in Synechocystis (9). Cryo-EM has been used to reconstruct a 3-dimensional (3D) model of the ϳ660 kDa BchID complex to 7.5 Å resolution (16). Subsequently, reconstructions of the ID complex bound to ADP, ATP, and the nonhydrolyzable ATP analog AMP-PNP have shown that the hydrolysis of ATP, which is essential for driving the chelation reaction, is accompanied by significant changes in the conformation of this complex (17).
In contrast with the structural detail available for the I and D subunits, less is known about the H subunit. Single-particle EM analysis of negatively stained BchH proteins from R. capsulatus revealed a three-lobed structure at a resolution of 25 Å. Differences in the structure were noted between the apo-and Protobound forms of H, and it was suggested that both the N-and C-terminal domains were important for porphyrin binding (18). The Mg chelatase BchH subunits from the purple photosynthetic bacteria such as R. capsulatus have sequence identities of 35-40% when compared with ChlH subunits from oxygen-evolving photosynthetic cyanobacteria, algae, and higher plants, but within this latter group sequence identities of over 60 -80% are common (supplemental Table S1). ChlH proteins, with 1326 amino acids and 148 kDa for the Thermosynechococcus elongatus protein, for example, are larger than their BchH counterparts (1189 residues and 129 kDa for the R. capsulatus protein). The alignment in supplemental Fig. S1 shows that one region in particular, marked in cyan, is responsible for this size difference. Possibly associated with this larger size, ChlH proteins appear to have acquired the capacity, during the course of evolution, to perform other functions. The Gun4 proteins of Arabidopsis and the cyanobacterium Synechocystis, for example, can form a complex with ChlH (19,20): Gun4 has a defined role in oxygenic phototrophs as an enhancer of the Mg chelatase reaction, acting as a molecular switch by reducing the Mg 2ϩ concentration required for activity at low porphyrin concentrations (21), and structural models are available for Gun4 from x-ray crystallography (21,22). In addition to possessing a binding site for Gun4, ChlH appears to have the capacity to interact with other proteins because several recent studies have shown that ChlH has functions extending well outside of the expected area of Chl biosynthesis. It was reported that the ChlH of higher plants is an abscisic acid receptor (23)(24)(25), although there is some disagreement in the literature on this point (26,27). In Synechocystis it has been proposed that ChlH binds to SigE, a positive regulator of carbohydrate metabolism (28).
In view of the structural and functional complexity of ChlH proteins, we undertook a single particle reconstruction study of ChlH using the recombinant protein from the thermophilic cyanobacterium T. elongatus, augmented by a small-angle x-ray scattering (SAXS) study of the highly homologous ChlH protein from Synechocystis. At the low resolutions obtained (ϳ30 Å), there is excellent agreement between the ChlH structures calculated using these differing approaches. ChlH is a large cage-like assembly adjoining a small, globular N-terminal domain. We discuss the possible roles of this ChlH structure during the catalytic cycle of Mg chelatase.

EXPERIMENTAL PROCEDURES
Cloning of T. elongatus chlH-The chlH gene was isolated from T. elongatus genomic DNA by PCR using Accuzyme reaction mix (Bioline) and the following primers: forward, 5Ј GCGCGAGCTCATATGTTCACCCACGTCAAGTCCACG 3Ј and reverse, 5Ј GCGCGGTACCGGATCCTTACTCAACCCC-TTCAATCTTGTCCTC 3Ј. Sequences complementary to the chlH gene are shown underlined. The forward primer contains SacI and NdeI restriction enzyme sites upstream of the ATG start codon, and the reverse primer contains BamHI and KpnI restriction enzyme sites downstream of the TAA termination codon. The PCR reaction was digested with NdeI and BamHI, which generated a 2.8-kb NdeI-BamHI upstream fragment and a 1.2-kb BamHI downstream fragment, as chlH carries an internal BamHI site. The upstream and downstream fragments were ligated successively into the pET9a-His 6 vector. The full-length chlH gene was checked by DNA sequencing prior to overexpression.
Purification of Recombinant ChlH from T. elongatus-The pET9a-His 6 ChlH plasmid was transformed into E. coli BL21(DE3) and cells were grown to an absorbance at 600 nm of 0.7, induced to overproduce recombinant protein with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside for 16 h at 25°C and harvested by centrifugation. The cell pellet was resuspended in immobilized metal affinity chromatography binding buffer (20 mM Tris-HCl (pH 7.4)/500 mM NaCl/45 mM imidazole) with the addition of 50 l of protease inhibitor mixture (Sigma). Cells were disrupted by sonication on ice (6 ϫ 15 s), and cell debris was removed by centrifugation for 20 min at 4°C. The cell-free supernatant was applied to a His60 Ni Superflow resin (Clontech) that was washed with immobilized metal affinity chromatography binding buffer until no eluted protein could be detected by its absorption at 280 nm. His 6 ChlH was eluted from the affinity column with a linear gradient from 0 -500 mM imidazole, and then exchanged into the binding buffer for ion exchange (50 mM Tricine-NaOH (pH 7.9)/0.3 M glycerol/1 mM DTT). The His 6 ChlH was then applied to a Q-Sepharose Fast Flow (Sigma) column and washed with ion exchange binding buffer until no eluted protein could be detected. His 6 ChlH protein was eluted using a 0 -500 mM NaCl gradient. The protein was further purified using a Biosep S4000 HPLC gel filtration column using gel filtration buffer (50 mM Tricine-NaOH (pH 7.9)/0.3 M NaCl/0.3 M glycerol/1 mM DTT). Sodium dodecyl sulfate polyacrylamide gel electrophoresis was performed as described previously (21).
Binding of D IX to ChlH-The ChlH-deuteroporphyrin IX (ChlH-D IX ) complex was analyzed using HPLC gel filtration. 1 M ChlH and 10 M D IX (above the K D of 1.48 M) were preincubated for 5 min prior to loading onto the HPLC column, and the absorption of eluate was monitored at 280 nm for ChlH and 398 nm for D IX . Tryptophan quenching assays were used to calculate the K D of ChlH for deuteroporphyrin (D IX ) as described previously (21).
Mg Chelatase Enzyme Assay-To ensure ChlH protein from T. elongatus was functional, the previously characterized Synechocystis Mg chelatase assay was used (21), but with the T. elongatus ChlH subunit substituted for the Synechocystis protein. Expression and purification of ChlH, ChlI, and ChlD subunits was performed as described previously (3). Formation of MgD IX product was monitored by its fluorescence at 590 nm, recorded using a Bio-Tek F2 microplate reader. The concentrations of Mg chelatase subunits used in the enzyme assays were ChlI, 0.2 M; ChlD, 0.1 M; and ChlH, 0.4 M.
Electron Microscopy-1 M ChlH was diluted to a suitable concentration, keeping the D IX concentration at 10 M to maintain the ChlH-D IX complex. 5 l of this solution was applied to carbon-coated 400-mesh copper grids (Agar Scientific) that had previously been glow-discharged for 30 s. Excess sample solution was removed by blotting the grid with filter paper. After twice washing with water, the grid was stained with 0.75% (w/v) uranyl formate for 20 s. Excess stain was removed by blotting and the grid was dried in air. Electron micrographs were recorded with a Philips CM100 microscope fitted with a 1K ϫ 1K Gatan Multiscan 794 charge-coupled device at ϫ61,000 magnification, resulting in a pixel size corresponding to 3.92 Å at the specimen level.
Image Processing-Single particles on digitalized micrographs were picked semiautomatically using EMAN2 (29) with a reticule size of 64 ϫ 64 pixels to accommodate selected particles while preventing any possible cut-off of structural information. Combined particle data were treated subsequently using the IMAGIC-5 software package (30). A band pass filter was applied to all particles to suppress low spatial frequencies (31). Unwanted background was removed by the use of a softedged circle mask. Masked data were normalized and centered for subsequent reference-free alignments to obtain averaged classes of the particles (32). A diverse set of characteristic classes was selected for multi-reference alignment calculation (31) and multivariate statistical analysis classification (33). This procedure was iterated until a stable classification result was obtained. An initial 3D model of the protein was determined by assigning Euler angles to the three most different class images using the angular-reconstitution program implemented in IMAGIC 5 with the "C1-start up" option. The model was refined by gradually adding input-averaged classes and by removing bad classes culminating in a total of 75 classes for apo-ChlH and 92 classes for substrate-bound ChlH. The Euler angles of data sets were refined against the anchor set produced from back projection of the previous models. This procedure was iterated by increasing the calculation precision and by decreasing the searching step size of the Euler angle until a stable model was obtained. These models were then used to produce a new reference that contained 480 projection images with uniformly distributed angles in Euler space for a new MRA calculation. Finally, 500 and 400 averaged classes were produced from 15910 and 12130 particles of apo-ChlH and porphyrin-bound ChlH samples, respectively. The final models were generated by rejection of 20% of the bad classes corre-sponding to 18.86% and 16.13% bad particles in the apo-ChlH and ChlH-D IX data sets, respectively.
The resolution of the final 3D model was estimated by calculating the Fourier shell correlation (FSC) between two 3D reconstruction models that were built separately from randomly selected halves of the two-dimensional projection data set. 0.5 Fourier shell correlations were used for resolution estimation (34).
Nanogold Labeling of ChlH-ChlH was buffer exchanged into nanogold binding buffer (20 mM Tris, 300 mM NaCl, 0.2% Tween 20, 5 mM imidazole). ChlH was added to nickel-nitrilotriacetic acid-Nanogold (Nanoprobes, Inc.) to give equimolar concentrations and was incubated at room temperature for 5 min. The solution was applied to an EM grid and imaged as described previously.
SAXS Sample Preparation and Data Collection-Purified Synechocystis ChlH protein was concentrated to appropriate concentrations in the range 1-10 mg/ml in chelatase buffer (50 mM MOPS/KOH (pH 7.7), 0.3 M glycerol, 1 mM DTT). Samples were prepared in a temperature-controlled brass cell with mica windows holding 100 l of volume and providing a x-ray path length of 1 mm. All data were collected at 4°C. The scattering data were collected at beam line 2.1 at the Daresbury Synchrotron Radiation Source (35). The data were calibrated using wet rat-tail collagen as well as silver behenate and reduced to onedimensional scattering profiles using beam line software (OTOKO, SRS Daresbury). Data were normalized with the incident radiation intensity using the photodiode reading in the beam stop, which also accounts for absorption properties of the sample. Low-and higher-angle scattering data were measured (4.5 m and 1 m sample-detector distance, respectively) at a wavelength of 1.54 Å with beam currents between 150 and 250 mA. When merged, the data from the two camera lengths (1 m and 4.5 m) covered the range of scattering vectors 0.01 Å Ϫ1 Ͻ q Ͻ 0.70 Å Ϫ1 (q ϭ 4sin/, where 2 is the scattering angle). Frames of 60 s exposure were integrated and averaged to a final total exposure time of 10 -60 min, depending on sample concentration. Frames at the beginning and end of each data collection were compared to exclude the possibility of protein aggregation or radiation damage during the course of data collection.
SAXS 3D Shape Reconstruction-Scattering profiles were treated with PRIMUS (36). The radius of gyration, R g , was evaluated using the Guinier approximation within PRIMUS for the low-angle scattering data and also with the indirect Fourier transform program, GNOM (37), using the entire scattering curve. GNOM also provided the maximum dimension, D max , defined by the distance r at which the distribution function p(r) approaches zero. GASBOR (38) was used to build structural models from the scattering data. Ten 3D shapes restored with GASBOR were superimposed and averaged with routines included in the DAMAVER suite of programs (39).

Purification and Functional Analysis of the ChlH Protein from T. elongatus-Recombinant
His-tagged ChlH from T. elongatus was purified as under "Experimental Procedures," with a final HPLC gel filtration step. Fig. 1A shows the elution profile from this final step, with detection of absorption at 280 nm. The main peak with a retention time of 38.0 min corresponds to a molecular mass of 150 kDa and therefore consists of monomeric ChlH. The small peak at ϳ35 min indicates the presence of a small population of ChlH dimers. The high degree of purity of the preparation was confirmed by the SDS-PAGE analysis shown in the inset in Fig. 1A. This preparation was used for analysis of negatively stained single particles of apo-ChlH.
This protein has not been produced or analyzed before, so some basic experiments were performed to demonstrate that the T. elongatus ChlH protein possesses the normal characteristics of its Synechocystis counterpart in terms of porphyrin binding and participation in a magnesium chelation reaction. To demonstrate that D IX could bind to ChlH forming a ChlH-D IX complex, samples of apo-ChlH (1 M) and D IX (10 M) were premixed and analyzed by HPLC gel filtration. In Fig. 1A, the elution traces correspond to detection at 280 nm (black line, protein absorption) and 398 nm (red line, D IX absorption). The apo-ChlH negative control elutes at 38 min with negligible absorption at 398 nm. In contrast, the premixed ChlH-D IX sample in Fig. 1A (bottom two traces) shows that the protein and porphyrin coelute, consistent with the formation of a monomeric ChlH-D IX complex. To quantify D IX binding, the apparent disassociation constant K D was determined using the method in Ref. 10, in which the progressive addition of D IX quenches fluorescence of Trp residues in the H subunit. Fig. 1B shows the plot of ChlH fluorescence as a function of D IX concentration. The curve fits a single-site binding model and yields a K D of 1.48 M in good agreement with the value of 1.22 M determined for the Synechocystis H subunit (10). Assays of Mg chelatase activity were performed by monitoring the formation of MgD IX product as described previously (21). The time course for the positive control, consisting of the H, I, and D subunits from Synechocystis, is shown in Fig. 1C (blue trace). There is no H subunit in the negative control (green trace). The formation of MgD IX in the red trace shows that the T. elongatus H subunit is able to substitute for the Synechocystis protein, although the chelation reaction proceeds at a slower rate. Thus, ChlH from T. elongatus is a Mg chelatase H-subunit.
Single Particle Analysis of Apo-ChlH and ChlH-D IX -Purified proteins were adsorbed onto freshly glow-discharged carbon-coated copper grids that were subsequently negatively stained. Fig. 2, A and D, shows the electron micrographs of apo-ChlH and ChlH-D IX samples, respectively. Under such conditions, about 70 particles with a clear background could be picked from a single 1K ϫ 1K micrograph. Fig. 2, B and E, show representative boxed images of single molecules of apo-ChlH and ChlH-D IX samples, respectively, with the presence of several domains within ChlH already evident from these images. Initial reference-free classifications were run on the data sets that contained 15930 particles of apo-ChlH and 12310 particles of the ChlH-D IX complex. The top rows of Fig. 2, C and F, show six selected averaged two-dimensional classes. Below are the 3D reconstructions viewed at the corresponding Euler angles, and the bottom rows show the corresponding reprojections from the 3D models, which are consistent with the averaged classes in each top row. An analysis of the distribution of Euler angles of the image classes used for the reconstitution of the  FEBRUARY 10, 2012 • VOLUME 287 • NUMBER 7 final models is presented in supplemental Fig. S2. In the case of both apo-ChlH and the ChlH-D IX complex there is a reasonable sampling of the orientations of these proteins in Euler space.

Structure of Cyanobacterial Magnesium Chelatase H Subunit
The resolution of the apo-ChlH and ChlH-D IX structures was calculated as 27 Å and 28 Å, respectively (supplemental Fig.  S3). The final 3D structural models are shown in Fig. 3, A and B, rotated successively by 90°about the z axis, and both filtered to 30 Å resolution for comparison. Apo-ChlH (Fig. 3A, cyan) shows a hollow structure with an internal cavity of ϳ100 nm 3 and a small globular domain connected to the rest of the protein by a narrow neck. Comparison with the ChlH-D IX complex (Fig. 3B, red) shows that there is little observable effect of substrate binding at this resolution. Fig. 3 also compares the 3D models of ChlH with the apo-and porphyrin-bound structures of BchH (C and D, respectively), taken from the earlier work of Sirijovsky et al. (18). The larger size of ChlH is apparent, mainly attributable to the region of ϳ100 residues indicated by the cyan bar in supplemental Fig.  S1.
To identify the globular "head" region in these reconstructions, we exploited the availability of the N-terminal His 6 tag and used 5 nm-diameter nitrilotriacetic acid-nanogold particles as a labeling reagent. Fig. 4 shows a gallery of labeled apo-ChlH proteins, with the electron-dense nanogold particles clearly visible in each case. In each image, ChlH appears to be joined to the nanogold by the globular head region, so we propose that this is the N-terminal domain.
SAXS Experiments on ChlH from Synechocystis-SAXS has become an important method for studying biological macromolecules, providing low resolution structural information on the molecular envelope of proteins in solution, and is therefore advantageous, as proteins can be studied in near-physiological conditions (40), providing valuable information that complements EM data on negatively stained single particles. The apo-ChlH from Synechocystis was used for SAXS analysis so the structural information obtained on a homogeneous and monodisperse solution of ChlH could be related directly to previous kinetic studies of Mg chelatase from Synechocystis (3, 7-13). The Synechocystis ChlH protein was prepared as described pre-viously (10), and the purity and monodispersity of the preparation were confirmed using SDS-PAGE and analytical HPLC gel filtration (data not shown). Fig. 5A shows the scattering pattern obtained for ChlH, and the inset, B, shows the p(r) plot. The scattering data allowed determination of the radius of gyration, R g , which gives an estimation of the distribution of mass within ChlH, defined as the root mean square difference of all atoms from the center of mass. The scattering curves also provided the maximum dimension, D max , defined by the distance r at which the distribution function p(r) approaches zero. The distribution function provides an estimation of the distribution of distances between scattering centers, again giving an idea of particle size. The R g was 46.9 Å Ϯ 0.5 Å and the D max value was 170 Å Ϯ 7 Å. To obtain a representative model for ChlH, the scattering data were analyzed with GASBOR (38). Ten GASBOR structures were superimposed and averaged, resulting in the model represented in Fig. 5, C and D, by a 3D matrix of hexagonally spaced closely packed atoms separated by 9 Å. The solution scattering   18. The apo-BchH structure is in cyan (C), and the porphyrin complex is in red (D). In the center and bottom rows, the molecules are rotated successively by 90°about the z axis. Scale bar ϭ 10 nm. The 3D models were generated using Chimera (64). data collected for Synechocystis apo-ChlH reveal a structure with a smaller domain attached to a larger one. SAXS data were also collected and analyzed for the ChlH-D IX complex, but no structural differences could be seen at the low resolution obtained (results not shown). The approximate dimensions of the ChlH solution structure shown in Fig. 5, C and D, are in good agreement with those obtained from the single particle reconstruction. To emphasize this close similarity, Fig. 6 shows a superposition of the apo-ChlH from T. elongatus from EM analysis and the SAXS structure of apo-ChlH from Synechocystis. The correspondence between these structures shows that the EM analysis on the basis of negatively stained single particles reflects the structure of the catalytically active protein in solution.
An N-terminal deletion of ChlH was constructed to investigate the structural requirements for Mg chelatase activity, its stimulation by Gun4, and the binding of D IX and MgD IX . The deletion was chosen for its strong likelihood of disrupting the "cage" structure of ChlH, so only 85.4 kDa of the original 148.6 kDa remain. The Synechocystis ChlH was used for these functional assays because the stimulatory effect of Gun4 is well documented for this protein (11,21,22). His-tagged recombinant ChlH proteins were purified (Fig. 7B) and assayed for their ability to bind D IX (OE) and MgD IX (‚) using quenching of Trp fluorescence as in Fig. 1B (see Fig. 7C). The K D values obtained were 5.66 Ϯ 0.74 and 3.51 Ϯ 0.68 M, respectively, which com-pared well with the K D values for the wild-type control of 4.37 Ϯ 0.33 and 5.28 Ϯ 0.76 M, respectively (data not shown), The lack of effect on porphyrin binding contrasts sharply with the complete abolition of Mg chelatase activity for the deletion mutant (see the 0 M Gun4 points in Fig. 7D). In these assays, the ChlH subunit variants were each combined with the ChlI and D subunits together with Mg 2ϩ , ATP, and D IX . The well documented stimulatory effect of Gun4 is seen for the wildtype enzyme (Fig. 7D, •). As an illustration of the ability of Gun4 to restore some activity to a previously inactive Mg chelatase, we included assays of the ChlH-gun5 mutant (Fig. 7D,  E), which harbors a single A942V mutation (11). However, when mixed with wild-type ChlI and D, the N-terminal deletion of ChlH is completely inactive, even with 2 M Gun4 present, which is a 5-fold molar excess over the ChlH concentration. This sizeable deletion of ChlH clearly distinguishes between the porphyrin-binding and catalytic functions of ChlH.  that requires 14 MgATP 2Ϫ (12). The transient state kinetic study by Viney et al. (13) showed that the key step in formation of a protein-bound MgProto product state is primarily the binding of MgATP 2Ϫ as opposed to its hydrolysis. These kinetic experiments did not monitor release of MgProto, so perhaps it is this process that is coupled to nucleotide hydrolysis, rather than insertion of Mg 2ϩ .

DISCUSSION
The SAXS and EM data show that ChlH forms an extended, asymmetric molecular assembly. The radius of gyration, R g , for this structure is 46.9 Å, similar to the R g value of 47.8 Å determined for ribulose 1,5-bisphosphate carboxylase/oxygenase (41), a compact, globular protein of much greater molecular mass, 534 kDa (42). Although R g values require a clear context to be interpretable in the form of a 3D molecular shape, this comparison supports the notion of ChlH as an open, asymmetrical structure.
The low resolution structural studies performed in this work cannot provide any direct mechanistic information, although it is interesting to note on the one hand the existence of a ϳ100 nm 3 cavity within the ChlH structure and on the other the likely need for an active site environment that promotes the Mg chelation reaction. This process has been studied in detail using density functional calculations, modeling the successive exchange of water molecules from the solvation shell surrounding the Mg 2ϩ ion with the pyrrole nitrogens of the Proto sub-strate, deprotonation of the Proto, and distortion of the porphyrin ring (43). It is possible that the ChlH protein helps to catalyze MgProto formation by exerting some control over the immediate solvation environment of porphyrin substrates and products, accommodating distortion of the porphyrin ring and facilitating the deprotonation of the pyrrole nitrogens. The theoretical study of Shen and Ryde (43) found that water is an unsuitable proton acceptor, which might necessitate the screening of that part of the active site involved in deprotonation from the bulk aqueous solvent. This presents a demanding requirement because Mg chelatase is a soluble complex located in the aqueous cytoplasmic compartment of the cell, or possibly at the interface between the membrane and cytoplasm.
The cage structure of ChlH is likely to have been severely disrupted by the N-terminal deletion of 566 residues (Fig. 7A). This deletion abolishes not only the catalytic activity of ChlH but also any restoration of activity by Gun4. It was shown previously that Gun4 restores partial activity to the A942V and P595L mutants of ChlH (11), the counterparts of A990V (gun5-1) and P642L (cch), respectively, in Arabidopsis ChlH (44). Nevertheless, binding of both D IX and MgD IX to the deletion mutant of ChlH is unaffected, showing that the catalytic function of ChlH requires the fully enclosed conformation and that porphyrin binding imposes far fewer demands on the structure. Given the likely necessity of controlling access of water to the active site, some form of enclosed structure for the H subunit is perhaps not surprising. As already noted by Sirijovski et al. (18), BchH proteins, such as that from R. capsulatus also have the potential to form an enclosing structure (see Fig. 3 for a comparison with ChlH). Single particle reconstruction of this protein showed that apo-BchH has three lobes, with the "thumb" and "finger" domains coming into contact upon binding of Proto (18). It was postulated that residues from both the N-and C-terminal regions are involved in binding Proto and that the majority of the porphyrin-binding residues are located within the N terminus. Proteolysis studies identified a flexible linker region, proposed to lie at the junction between the Nand C-terminal domains and flanking Gly-734 (18).
The apo-ChlH and ChlH-D IX structures, both filtered to 30 Å for ease of comparison in Fig. 3, do not show any evidence of large-scale alterations in conformation upon porphyrin binding, and a more detailed structural analysis is required to examine this point further.
The one feature of ChlH of T. elongatus that can be identified is the N-terminal domain. This is the "head" region, assigned on the basis of labeling with a nitrilotriacetic acid-nanogold particle that binds to the N-terminal His 6 tag (Fig. 4). From estimates of the fraction of ChlH represented by the head domain, it has a molecular mass of 15.7 kDa, corresponding to ϳ152 amino acids at the N terminus followed by a short linker of a few residues. The large error in this estimation gives a range of 14.1-17.3 kDa for the molecular mass of this domain, which corresponds to the sequences running from the N terminus to either Gly-127 or Phe-156. In this region ChlH has the following sequence: G 127 SFSLAQIG 135 QSKSVIANFMKKRKEKSG-153 AG 155 F. We note that Gly-127 is conserved in all Mg chelatase H subunits, and a counterpart is also found in the CobN cobaltochelatase subunit from Pseudomonas denitrificans. Although the positions of the other Gly residues vary slightly, ChlH sequences from diverse organisms contain in this region, in addition to a Gly-127 equivalent, two to four other Gly residues. Although the overall isoelectric point for ChlH is 5.30, the sequence between Gly-127 and Phe-156 has a pI of 10.6. Such contrasting values are predicted for other ChlH sequences, and it is possible that this Gly-127-Phe-156 region controls the binding of ChlH to other subunits or to the membrane or is involved in mobility of the head domain, possibly opening or closing the cavity within ChlH. The lumen enclosed by ChlH is ϳ100 nm 3 in vol-ume, equivalent to a sphere 4.7 nm in diameter and much larger than required to sequester a Proto molecule of 1.4 ϫ 1.4 nm. This discrepancy could arise from a pooling of negative stain in this cavity, which would lead to an overestimation of the enclosed volume, as seen in an EM and x-ray crystallographic analysis of the glycerol dehydrogenase complex (45). It is interesting to note that the hole adjacent to the flexible head domain is easily large enough to allow access of Proto to the lumen or exit of the MgProto product, so it is possible that the movement of the head domain of ChlH plays an important role in the catalytic cycle. . Superposition of the 3D reconstruction of apo-ChlH from T. elongatus (cyan) from EM analysis and the apo-ChlH from Synechocystis (yellow) from the SAXS analysis. The superposition is viewed from three angles to emphasize the similarity between the two structural models. The handedness of the 3D reconstruction was selected arbitrarily for the best fit.

Multiple Roles of the Mg chelatase H Subunit in BChl
Biosynthesis-The multiple functions for the Mg chelatase H subunit could well necessitate a large, multidomain protein.
Apart from its main role as the porphyrin-binding subunit of Mg chelatase, the H subunit is likely to play a regulatory role by virtue of its position at the branch point between heme and BChl biosynthesis. The need for regulatory mechanisms is clear because the levels of these pigments have to be carefully metered to ensure that they are neither deficient, which would impair assembly and repair of the photosynthetic apparatus, nor overproduced, which could result in light-induced photodamage to the cell. The metabolic versatility of some photosynthetic bacteria imposes an additional regulatory load on the cell because several species can shift between aerobic/respiratory and anaerobic/photosynthetic growth modes, which requires a reversible switching of flux down the heme and Bchl pathways. As an example, Willows et al. (46) demonstrated that BchH is inactivated by light and oxygen as cells of R. capsulatus switch to aerobic growth (46). Furthermore, BchH increases the activity of the next enzyme in the Bchl pathway, S-adenosyl-L-methionine (magnesium protoporphyrin IX methyltransferase) (47,48), although this stimulation was not found in other work (49). BchH of R. capsulatus forms a complex with either BchM or BchJ proteins (50). Finally, green sulfur bacteria such as Chlorobaculum tepidum possess three homologs of BchH, designated as BchH, S, and T, no doubt linked to the more complicated pigment biosynthetic pathways in such bacteria, which lead to Chl a, Bchl a, and Bchl c (51,52). It was proposed that the three BchH homologs apportion the amounts of each pigment produced by the cell (52). A detailed in vitro enzymological study of these C. tepidum BchH homologs, combined with BchI and D, showed that their Mg chelatase activity varied over 5 orders of magnitude. Moreover, two of the BchH homologs increased activity of the MgProto methyltransferase, but one of them decreased it (48).
Multiple Roles of the Magnesium Chelatase H Subunit in Chl Biosynthesis-ChlH in oxygenic photosynthetic organisms does not have to contend with multiple Chl/Bchl pathways in the same cell as in Chlorobaculum nor switches between aerobic/respiratory and anaerobic/photosynthetic growth as in Rhodobacter, but the task of regulating flux down the Chl pathway remains, particularly in view of the fluctuating demands on the pathway placed by diurnal rhythms, variations in light intensity (53,54), and repair of damaged photosystem II complexes (see Ref. 55 for a review of Chl biosynthesis and its regulation). A regulatory role for ChlH in higher plants was proposed several years ago (53,56), and it is established that ChlH stimulates activity of Mg Proto methyltransferase in Synechocystis (57)(58)(59) and in tobacco (60). As already noted, the Gun4 protein in higher plants and cyanobacteria can form a complex with ChlH (19,61,62), a process that heavily influences the activity of Mg chelatase (21). A wider role for Gun4 has been proposed, including regulation of 5-aminolevulinic acid synthesis and in photoprotection (63). It is possible that these multiple regulatory and catalytic functions could account for the increased size and structural complexity of ChlH with respect to BchH apparent in the comparison of these structures in Fig. 3. The extra size and complexity of ChlH might confer properties on ChlH subunits from oxygenic photosynthetic organisms that are absent in BchH subunits.
We propose that ChlH has adopted a caged structure, in comparison with BchH, as a response to the intracellular environment of Mg chelatase in T. elongatus, an oxygen-evolving photosynthetic organism. In view of the lability of MgProto, we suggest that ChlH encloses this product of magnesium chelation and chaperones it to the active site of the methyltransferase to ensure both efficient handover of MgProto and its protection from photooxidation. Given the sequence homology of the T. elongatus ChlH to the Mg chelatase H subunits of higher plants and algae, it is likely that an enclosed structure is required to discharge the catalytic, chaperoning, and regulatory functions of this H subunit in all oxygenic photosynthetic organisms.