N1,N1-Diethyl-N 2-(2,3,4,6-tetra-O-acetyl-β-D-glucopyranosyl)acetamidine

Acta Cryst. (2003). D59, 2183±2190 Xu et al. PII proteins 2183 research papers Acta Crystallographica Section D Biological Crystallography ISSN 0907-4449 The structures of the PII proteins from the cyanobacteria Synechococcus sp. PCC 7942 and Synechocystis sp. PCC 6803 Yibin Xu, a ² Paul D. Carr, b ² Paula Clancy, a Mario Garcia- Dominguez, b,c Karl Forchhammer, d Francisco Florencio, c Nicole Tandeau de Marsac, e Subhash G. Vasudevan a,f and David L. Ollis b * a Department of Biochemistry and Molecular Biology, James Cook University, Townsville, Queensland 4811, Australia, b Research School of Chemistry, Australian National University, GPO Box 414, Canberra, ACT 2601, Australia, c Instituto de Bioquõ Âmica Vegetal y Fotosõ Ântesis, Universidad de Sevilla-Consejo Superior de Investigaciones Cientõ Âficas, Ame Ârico Vespucio s/n, E-41092 Sevilla, Spain, d Institut fu Èr Mikrobiologie und Molekularbiologie, Justus-Liebig-Universita Èt Giessen, Germany, e Unite Âdes Cyanobacte Âries, CNRS-URA 2172, De Âpartement de Microbiologie Fondamentale et Me Âdicale, Institut Pasteur, 28 Rue du Docteur Roux, 75724 Paris CEDEX 15, France, and f Novaris Institute for Tropical Diseases Pte Ltd, 1 Science Park Road, #04-14 The Capricorn, Singapore Science Park II, Singapore 117528, Singapore ² These two authors contributed equally to this paper. Correspondence e-mail: [email protected]u #2003 International Union of Crystallography Printed in Denmark ± all rights reserved The PII proteins from the cyanobacteria Synechococcus sp. PCC 7942 and Synechocystis sp. PCC 6803 have been crystallized and high-resolution structures have been obtained using X-ray crystallography. The core of these new structures is similar to that of the PII proteins from Escherichia coli, although the structures of the T- and C-loops differ. The T-loop of the Synechococcus protein is ordered, but appears to be stabilized by crystal contacts. The same loop in the Synechocystis protein is disordered. The C-terminus of the Synechocystis protein is stabilized by hydrogen bonding to the same region of a crystallographically related molecule. The same terminus in the Synechococcus protein is stabilized by coordination with a metal ion. These observations are consistent with the idea that both the T-loop and the C-terminus of PII proteins are ¯exible in solution and that this ¯exibility may be important for receptor recognition. Sequence comparisons are used to identify regions of the sequence unique to the cyanobacteria. Received 26 May 2003 Accepted 5 September 2003 1. Introduction PII is a signal-transduction protein that is best known for its role in the regulation of nitrogen assimilation by enteric bacteria. However, the protein is found and highly conserved in a variety of organisms covering all the three kingdoms of life (Ninfa & Atkinson, 2000; Arcondeguy et al., 2001). In several of these organisms there are genes that code for two or even three PII-like proteins. In Escherichia coli there are two such proteins, known as GlnB and GlnK. Whilst these two proteins share 67% similarity, they do not appear to be redundant in their function (van Heeswijk et al., 1995; Atkinson & Ninfa, 1999). The role of the PII-like proteins in other organisms is only now being elucidated, but it is already clear that they have different functions. In E. coli, the PII paralogues are uridylylated and modulate the activity of glutamine synthetase (GS) and the transcription of the corresponding gene glnA (Merrick & Edwards, 1995). In cyanobacteria, the PII protein is reversibly phosphorylated on a seryl residue (Forchhammer & de Marsac, 1995a). The speci®c phospho-PII phosphatase, a member of the PP2C family, was identi®ed recently (Irmler & Forchhammer, 2001). The cyanobacterial PII protein is involved in the regulation of nitrate and nitrite uptake (Forchhammer & de Marsac, 1995b; Lee et al., 1998, 2000), bicarbonate transport (Hisbergues et al., 1999) and is required for the activation of NtcA-dependent gene expression under conditions of nitrogen starvation (Aldehni et al., 2003). The structures of E. coli GlnB (EcPII) and GlnK have both been solved (Cheah et al., 1994; Xu et al., 1998). More recently, research papers 2184 Xu et al. PII proteins Acta Cryst. (2003). D59, 2183±2190 the structure of the PII protein from Herbaspirillum seropedicae (HsPII), a member of the same order (proteobacteria) but a different subgroup to E. coli, has been solved (Benelli et al., 2002). In all three cases, signal transduction occurs through modi®cation of the Tyr51 residue found on the T-loop which extends out from the core of the molecule. This loop has been found to have different conformations in the E. coli proteins and is disordered in the complexes formed with ATP (Xu et al., 1998, 2001). The T-loop is also largely disordered in HsPII. It is thought that this loop is ¯exible in solution and that this ¯exibility allows recognition to occur with several receptors. Apart from the T-loop, the C-terminus of the PII protein appears to be capable of taking on multiple conformations and may be involved in recognition. The C- terminal loop adopts one of two topologies: either a pair of short antiparallel -strands or a 3 10 -helix. This loop is observed to exhibit rigid-body movements relative to the core of the molecule when known structures are overlaid. The E. coli proteins require ATP in order to carry out their functions. Crystallographic studies have shown that ATP binds in a highly conserved cleft on the side of the molecule. Like their E. coli counterparts, the PII proteins from cyanobacteria form trimers of 12 kDa subunits. For Synechococcus sp. PCC 7942, it has been shown that the PII protein (SnPII) can form with zero, one, two or three phosphorylated serine residues. The extent of phosphorylation depends upon the availability of carbon and nitrogen sources (Forchhammer & de Marsac, 1995b; Lee et al., 1998). At the molecular level, the phosphorylation state of PII responds to central carbon metabolites, in particular 2-oxoglutarate, re¯ecting the metabolic state of the cells (Ruppert et al., 2002). The serine residue in question is located in the T-loop at position 49 (Forchhammer & de Marsac, 1995a). When expressed in E. coli, the tyrosine at position 51 can be uridylylated, but it does not modulate either GS activity or the transcription of its gene (Forchhammer, 2003). This suggests that the T-loop in SnPII differs from that found in the E. coli paralogues. Similarly, studies on PII from Synechocystis sp. PCC 6803 (SsPII) have shown that the same seryl residue at position 49 is phosphorylated (Garcia-Dominguez & Florencio, unpublished work). The sequences of the T-loops in the SnPII and SsPII proteins are identical and differ from that of EcPII at 5 of the 19 residues (Fig. 1). The sequence identities between the SnPII and SsPII proteins and the EcPII protein are 65 and 55%, respectively. This compares to 86% identity between the two cyanobacterial sequences. Although the sequences of the cyanobacterial PII proteins are very similar to those of the E. coli PII-like proteins, they are differently modi®ed and have different target receptors. The cyanobacterial proteins are more closely related to the PII proteins found in plants. The sequence of the PII from Arabidopsis thaliana is more similar to that of the cyanobacteria than it is to E. coli (Smith et al., 2003). In addition, the plant proteins do not have a tyrosine in the T-loop but, like the cyanobacterial proteins, have a serine that appears to be phosphorylated. We have determined the structures of the SsPII protein and a mutant form of the SnPII protein. The mutant had an alanine in place of the functionally important Ser49 and mimics the unphosphorylated form of the protein (Lee et al., 2000). We comment on the functional signi®cance of structural comparisons made between the PII proteins from E. coli and cyanobacteria. 2. Materials and methods 2.1. Bacterial strains and growth conditions E. coli DH5was used for plasmid construction and replication. E. coli BL21 was used for the expression of SsPII protein from Synechocystis sp. PCC 6803 (hereafter referred to as Synechocystis). The vector expressing the mutant form of the PII protein (SnPIIA) from Synechococcus sp. PCC 7942 (hereafter referred to as Synechococcus) was transformed into RB9065 cells. E. coli strains were grown in Luria Broth (LB) medium as described by Sambrook et al. (1989) supplemented with 100 mgml ÿ1 ampicillin when required. 2.2. Expression and purification For expression of the SsPII protein in E. coli, a plasmid, pMAB11, was constructed as a derivative of the expression vector pET-3a (Novagen). A DNA fragment of about 400 bp encompassing the Synechocystis glnB gene was obtained by PCR ampli®cation with oligonucleotides gbp1 (50-GTACCA- CATATGAAAAAAG-30) and gbp2 (50-CTTGTCTGGATC- CGCCCAAC-30). Oligonucleotides gbp1 and gbp2 harbour NdeIandBamHI sites, respectively. The PCR product was checked by sequencing, digested with NdeI and BamHI and cloned into the NdeI±BamHI sites of pET-3a. Oligonucleotide gbp1 substitutes the natural translation starting codon of the Synechocystis glnB gene (TTGl; Garcia-Dominguez & Florencio, 1997) by a standard ATG codon, included in the NdeI site. pMAB11-transformed E. coli BL21 strain was grown in 3 l of M9 minimal medium with ampicillin (100 mgml ÿ1 ) to an optical density at 600 nm of 0.5; for production of the SsPII protein, isopropyl--d-thiogalactopyranoside (IPTG) was then added to a ®nal concentration of 1 mMand the culture Figure 1 Alignment of two cyanobacterial PII sequences with EcPII. was further incubated for 4 h before harvesting cells by centrifugation. Cells were resuspended in 15 ml T buffer (50 mMTris±HCl pH 8.0, 50 mMNaCl, 5 mMMgCl 2 ,1mMEDTA), disrupted by sonication (20 kHz, 75 W) for 2 min and centrifuged at 40 000 gfor 30 min. Cell-free extract was fractionated between 35 and 50% ammonium sulfate. After dialysis with T buffer, the protein was heated at 338 K for 5 min and centrifuged; the supernatant was then applied to a DEAE-cellulose column. The ¯owthrough was collected, concentrated with 50% ammonium sulfate and desalted in a Sephadex G-50 (Pharmacia) column equilibrated in T buffer. A 150 mg fraction of SsPII protein was yielded and adjusted to 10 mg ml ÿ1 with T buffer. The gene for the SnPIIA protein was obtained from plasmid pPM308 (Lee et al., 2000). This plasmid was used to produce a PCR fragment with an NdeI and KpnI site at either end. The PCR primers used for this purpose were SynPII-FP (50- TATATTCATATGAAGAAGATTGAGGCGATTATTC-30) and SynPII-RP (50-TATATTGGTACCTTAGATCGTGTCG- GCGTTTTTC-30). The PCR product was digested with NdeI and KpnI restriction endonucleases and ligated with pND707 (Love et al., 1996) that had been similarly cut. The resulting clone was checked by nucleotide-sequence analysis. The SnPIIA protein was overexpressed in RB9065 cells (UTase ÿ mutant). The cells were grown in a 5 l fermenter of LB medium with ampicillin (100 mgml ÿ1 )at303KtoanOD 595 of 0.6; the temperature was then rapidly shifted to 315 K and the cells were grown for a further 3 h. The cells were harvested by centrifugation, resuspended in HEPES buffer A(20 mM HEPES, 1 mMEDTA, 1 mM-mercaptoethanol) and disrupted using a French press. The cells were pressed twice and PMSF (0.5 mM) was added before and after pressing to prevent proteolysis. Nucleic acid contaminants were removed from the sample using streptomycin sulfate [1.5%(w/v)] precipitation, followed by fractionation using ammonium sulfate (between 25 and 35%). The sample was dialysed against 3 2 l changes of HEPES buffer Aand applied to a Cibacron Blue matrix column, washed with the same buffer containing 100 mMNaCl and ®nally eluted with a salt gradient in HEPES buffer (1±0 MNaCl, 5 mMATP). The resulting protein solution was >95% in purity based on a 15% SDS± PAGE gel and was concentrated to 20 mg ml ÿ1 for crystallization trials. 2.3. Crystallization, data collection and structure determination Crystals of the SsPII protein were grown using the hangingdrop method at 293 K. Equal volumes of the protein solution (10 mg ml ÿ1 ) and a reservoir solution containing 20% 2-propanol, 100 mMCaCl 2 , 100 mMsodium acetate pH 4.7 were used. Small crystals appeared after 3 d and continued to grow for about two weeks. Crystals were prepared for data collection by transfer to a solution of the reservoir buffer containing 25% glycerol. The SnPIIA protein was crystallized using the hanging-drop method at 293 K by mixing equal volumes of the protein solution (20 mg ml ÿ1 ) with a reservoir solution containing 1.0 Mlithium sulfate, 0.1 MTris buffer pH 8.5 and 10 mM nickel chloride hexahydrate. Crystals were prepared for data collection by transfer to a solution of the reservoir buffer. All data were collected using a Rigaku R-AXIS IIc image plate mounted on a rotating-anode generator running at 50 kV, 100 mA. Data were processed with the HKL program (Otwinowski & Minor, 1997). For SsPII, data were collected at 100 K. The crystals were found to belong to space group R3, with unit-cell parameters a=b= 129.79, c= 74.14 A Ê, and contained four subunits in the asymmetric unit. For the SnPII S49A mutant, data were collected at 277 K. The crystals were found to belong to space group I4 1 22, with unit-cell parameters a=b= 109.70, c= 109.68 A Ê, and contained a trimer in the asymmetric unit. The structure of the SsPII protein was determined by molecular replacement using a model of a trimer of EcPII (Carr et al., 1996) from which the T-loop (residues 37±55) and the four C-terminal residues had been deleted. Calculations were undertaken using the program AMoRe (Navaza, 1994) as implemented in the CCP4 package (Collaborative Computational Project, Number 4, 1994). The best rotation-function solution had a correlation coef®cient of 0.424 compared with the highest incorrect peak of 0.170. The best translationfunction solution had a correlation coef®cient of 0.509 and an Rfactor of 0.465 compared with 0.237 and 0.566, respectively, for the highest incorrect solution. After rigid-body re®nement, these values improved to 0.587 and 0.428, respectively. The model generated from this solution gave good crystal packing when inspected on a graphics terminal. Further re®nements were undertaken using CNS (Bru Ènger et al., 1998) and phases were improved using DM (Cowtan, 1994). During subsequent model building, it became apparent that there was density for another molecule of PII in the asymmetric unit. A further monomer of PII was added to the model. A well packed molecule was obtained from four subunits of PII. The original three molecules formed a trimer around a noncrystallographic triad, whilst the fourth produced trimers from crystallographically related molecules. A Matthews coef®cient (Matthews, 1968) V M of 2.8 A Ê 3 Da ÿ1 is obtained for four molecules per asymmetric unit, with a corresponding solvent content of 56%. This compares with a V M of 3.6 A Ê 3 Da ÿ1 for a single trimer per asymmetric unit or 1.8 A Ê 3 Da ÿ1 for two trimers per asymmetric unit. Further rounds of re®nement using simulated-annealing, positional and individual B-factor re®nement using standard CNS protocols (Bru Ènger et al., 1998) were interspersed with manual rebuilding of the model. The ®nal model had an R work and R free of 0.219 and 0.251, respectively. Re®nement statistics are listed in Table 1. The structure of the SnPIIA protein was solved by molecular replacement with MOLREP (Vagin & Teplyakov, 2000), incorporated in the CCP4 suite. The high-resolution structure of the E. coli PII trimer without the T-loop was used as a search model (Carr et al., 1996). MOLREP gave rise to a Acta Cryst. (2003). D59, 2183±2190 Xu et al. PII proteins 2185 research papers research papers 2186 Xu et al. PII proteins Acta Cryst. (2003). D59, 2183±2190 dominant solution with R= 0.481 and a correlation coef®cient of 0.476 to 3.0 A Êresolution. Rigid-body re®nement and further structure re®nement were implemented using CNS v.1.1 and model building was carried out with O(Jones et al., 1991). Non-crystallographic symmetry restraints were applied throughout the re®nement. The occupancies of disordered side chains were set to zero during the re®nement. A number of non-peptide features appeared in the electron-density map during the course of re®nement. Within each ATP-binding site, there were two features that appeared to be sulfates from the crystallization buffer. In addition, there was a spherical feature in the maps that appeared to be Ni 2+ ions, also from the crystallization buffer. Statistics of data collection and structure re®nement are given in Table 1. 3. Results and discussion 3.1. The models The models con®rm that the cyanobacterial PII proteins have the same overall topology as other known PII structures. The molecules form homotrimers around threefold axes, with -sheet interactions involving both of the neighbouring protomers. These hydrogen bonds hold the central core structure of the trimers together. Each monomer consists of a central four-stranded antiparallel -sheet with a pair of antiparallel helices on one side. These form an interlocking double  motif. There are two major loops emanating from the central sheet. The ®rst and largest, between strands 2 and 3, is known as the T-loop and contains the site of phosphorylation. In structures solved to date, this loop has only been visible where crystal-packing forces have stabilized one particular conformation. This has led to the conclusion that it is ¯exible in solution. The T-loop is not visible in the SsPII structure reported here, but is visible in the electron density of SnPII (Fig. 5). Lattice contacts appear to stabilize this conformation. The second loop is found between strand 4 and the carboxylterminus. In both of the structures reported here and in several other PII structures this loop contains two short antiparallel -strands. In other PII-like structures, for example, EcGlnK and HsPII, this loop contains a 3 10 -helix rather than -strands. The model of SsPII consists of four protein chains. Chains A,Band Cform a trimer around a non-crystallographic triad. Chain Dforms a trimer by symmetry operations on itself. Like several other forms of PII (Xu et al., 2001), the T-loop (residues 37±53) is not visible in the electron density, presumably owing to mobility. The model consists of residues A1±36, A54± 112, B1±36, B55±112, C1±36, C55±112, D1±38 and D54±110. In addition, two Ca 2+ ions, two glycerol molecules and 217 water molecules were included. The following residues did not exhibit good side-chain density and were truncated to alanines in the re®nement: A76, A101, A106, B76, C76, D38 and D54. The following glutamate residues exhibited alternate conformations of their side chains: A15, A62, B5, B15, B62, C5, C15 and D15. Alternate conformers were modeled at half-occu- pancy. The ®nal model exhibited good stereochemistry when checked by WHAT_CHECK (Hooft et al., 1996) and PROCHECK (Laskowski et al., 1993). The Ramachandran plot showed 98.1% of residues falling into the most favoured region, whilst the remaining 1.9% were in additional allowed regions. No residues fell into the disallowed or generously allowed regions. The overall fold of the SsPII protein is essentially the same as that of EcPII protein. The individual protein chains were overlaid using LSQKAB (Kabsch, 1976). The r.m.s. values for main-chain atom displacement were less than or equal to 0.6 A Ê. For comparison, the Dchain was also superimposed on Table 1 Summary of X-ray diffraction data and re®nement statistics. Values in parentheses are for the outer data shell. SnPII SsPII X-ray diffraction data Resolution (A Ê) 25±2.0 (2.1±2.0) 25±2.0 (2.1±2.0) Observed re¯ections 68888 156138 Unique re¯ections 19247 31543 Completeness (%) 84.0 (80.2) 99.5 (98.2) R merge (%) 5.9 (24.1) 5.3 (16.7) Average I/(I) 8.5 (3.1) 16.7 (4.93) Re®nement statistics Resolution (A Ê) 25±2.0 25±2.0 No. of re¯ections (working) 17030 29381 No. of re¯ections (test) 1869 (8.2%) 1577 (5%) R cryst /R free ² (%) 20.0/22.7 21.9/25.3 No. protein atoms 2610 3499 No. solvent atoms 132 217 No. glycerol molecules Ð 2 No. ions 3 Ni 2+ ,6SO 2ÿ 42Ca 2+ R.m.s. deviation from ideal geometry Bond length (A Ê) 0.005 0.006 Bond angle () 1.28 1.26 Dihedrals () 23.64 26.64 Improper () 0.67 1.17 Average Bfactors (A Ê 2 ) Protein 18.4 29.8 Water 30.0 42.3 Sulfate 43.4 Ð Glycerol Ð 51.2 Ions 23.7 34.2 Ramachandran plot (%) Most favoured regions 96.9 98.1 Additional allowed regions 3.1 1.9 Generously allowed regions 0.0 0.0 Disallowed regions 0.0 0.0 ²R cryst =(|F obs |ÿ|F calc |)/|F obs |; R free was calculated in the same manner as R cryst using a test set of data not used in the re®nement process. Table 2 R.m.s. displacement values of superposition of main-chain atoms of various PII structures. Species Space group Chain R.m.s. (A Ê) Synechocystis R3A0.60 B0.59 C0.58 D0.00 Synechococcus I4 1 22 1.24 E. coli P6 3 0.95 I2 1 3 1.19 E. coli GlnK P2 1 3 1.08 H. seroedicae P2 1 2 1 2 1 A0.92 a number of other PII structures (Table 2). In all of the superpositions the basic structure of the protein chains align well, with just three regions showing variations. Firstly, in structures where crystal packing stabilizes the T-loop (EcPII, EcGlnK and SnPIIA) it adopts different conformations. Secondly, the C-loop, which exhibits -sheet topology, undergoes small rigid-body movements with respect to the central strands. Thirdly, the last three C-terminal residues differ between structures. Fig. 2 shows an overlay of the various PII protein chains. Inspection of the temperature factors reveals that the central rigid core of interlocking -strands has the lowest Bfactors (18 A Ê 2 ). The B-loop and helices have values that are typically double those of the core. The C-loop, however, is more stable (25 A Ê 2 ) owing to inter-trimer strand formation. There is variation between the individual protomers that cluster around the non-crystallographic threefold axis, although it dif®cult to see why this should be as they have similar packing environments. The crystal packing appears to have in¯uenced the current structure by interactions at two sites. The ®rst is at the Ca 2+ ion-binding sites, which occur between symmetry-related molecules. The metal ion is liganded to the acidic side chains of residues Asp14 and Glu15 and the carbonyl O atom of Phe11. It is also bound to the symmetry mates of these three atoms in a neighbouring molecule, giving the Ca 2+ ion full octahedral coordination (Fig. 3). The Acta Cryst. (2003). D59, 2183±2190 Xu et al. PII proteins 2187 research papers Figure 2 Superposition of various PII monomers. Colour scheme: SnPIIA, magenta; SsPIIA, blue; SsPIID, orange; EcPII, black; GlnKB, green; HsPII, red. Note that T-loops (upper left), where present, show variable conformations. The C-loop structure (lower right) undergoes rigid-body movements relative to the stable double  core between structures. Figure 3 Close-up of crystal contact in SsPII at the Ca 2+ -binding site. Colour coded by (symmetry-related) protein chains. Figure 4 Hydrogen bonds stabilize the C-loops of symmetry-related molecules of SsPII. research papers 2188 Xu et al. PII proteins Acta Cryst. (2003). D59, 2183±2190 second site of packing-in¯uenced interactions occurs at the ®nal -strand, which forms part of the C-loop. Here, the backbone amide N and carbonyl O atoms of residues 107 and 109 form hydrogen bonds in a classic antiparallel -sheet arrangement with the same residues in a neighbouring symmetry-related molecule (Fig. 4). This has the effect of producing a large extended -sheet of 12 strands spanning two trimers. Similar packing interactions were observed in the EcPII I2 1 3 structure (Xu et al., 2001). The model of the SnPIIA protein consists of a trimer of almost identical subunits. The T-loops are clearly ordered as shown in Fig. 5. The main interaction stabilizing the T-loops appears to be main-chain hydrogen bonds forming -sheets involving residues 49±53 with their symmetry-equivalent residues from another trimer. Although Ser49 has been mutated to alanine in this structure, it does not appear that side-chain interactions involving the serine would affect the local conformation of the loop. The side chains of residues Arg38, Gln39, Lys40, Gln42 Arg47, Glu50, Gln57, Gln69 and Glu85 in all three subunits have poorly de®ned electron density. The C-terminal loop has -sheet topology. There are three Ni 2+ ions that link the N- and C-termini of each subunit, as shown in Fig. 6. In addition, there are 132 solvent molecules and six sulfates that bind in the three ATP-binding clefts. PROCHECK (Laskowski et al., 1993) indicates that 96.7% of the ', angles of all non-glycyl residues are in the most favoured regions of the Ramachandran plot. The additional 3.3% were in the additionally allowed regions. Inspection of the Bfactors again shows the central core of interlocking -strands to have the lowest mobility (12 A Ê 2 ). All loops, including the T-, B- and C-loops, have values that are approximately doubled. The external helices are intermediate between these two values. The nickel ions have B factors similar to the liganding residues. The SO2ÿ 4ions have higher Bfactors than the nickel ions, closer to typical solvent values. 3.2. Molecular structures The tertiary structures of the cyanobacterial SsPII and SnPIIA proteins are essentially the same as the EcPII and GlnK structures. However, there are signi®cant differences between the T- and C-loop regions of SsPII and SnPII and the same loops in EcPII and GlnK. The T-loop of the SsPII protein is disordered as has been observed in some EcPII and GlnK structures. The T-loop in the SnPIIA protein extends from 2 and 3 without a disruption of secondary structure, so that these strands are longer than those found in EcPII and GlnK. 2 of SnPIIA consists of residues 29±41, this being ®ve more than the number of residues in 2 (37± 41) of the EcPII molecule. Similarly, 3of SnPIIA consists of residues 50±65, while 3 of the other PII molecules consists of residues 56±65. There is little contact between the extension of 2 and 3 and the core of the Synechococcus PII molecule. The T-loop region of SnPIIA takes part in crystal packing. Residues 49, 51 and 53 in a given monomer of the trimer form hydrogen bonds with their symmetry counterparts of one monomer within a neighbouring trimer. The speci®c interactions involve backbone hydrogen bonds between Tyr51 and its crystallographic equivalent, while Ala49 and Val53 are linked to Val53 and Ala49, respectively, in a crystallographically related molecule. These interactions appear to stabilize the T-loop. Figure 5 Trimers of PII with visible T-loops. All are stabilized by crystal contacts. Colour scheme: SnPII, magenta; GlnK, green; EcPII, black. (a) Left, view looking down the threefold axis. Right, perpendicular to the threefold axis looking into the ATP-binding cleft. (b) An omit map of the T-loop region in SnPII. The C-terminus of the SsPII protein is very similar to that observed in a low-pH structure of EcPII (Xu et al., 2001). In both, the C-terminus is involved in crystal contacts. The conformation of the C-terminus of SsPII differs from that observed in EcPII and GlnK structures obtained at higher pH values. The conformation of the C-terminus of SnPIIA differs from those of other PII proteins. It points towards the N- terminus of the same chain and is held in place by coordination with a metal ion, presumably a Ni 2+ that was added for crystallization. The coordination site involves Asp110, a residue that is not conserved in either the SsPII or the EcPII proteins. The metal-ion coordination ties down the C-terminus and probably facilitates crystallization, but is not likely to have a physiological function. SnPIIA binds two anions in the ATP-binding cleft. In the absence of ATP, GlnK was also observed to bind an anion in this cleft. The GlnK anion-binding site also appears to be occupied in SnPIIA. This anion forms interactions with the backbone N atom of Gly87 and the side-chain N atoms of Arg101 and Arg103 from a neighbouring monomer within the trimer. The second anion forms links to the side-chain N atoms of Lys90 and the neighbouring Arg101 as well as the symmetry-related Arg103. The latter anion is exposed to solvent and has a higher Bfactor. 4. Concluding remarks The structures observed for the cyanobacterial PII proteins are very similar to those observed for the E. coli proteins. The core of the molecule is remarkably well preserved; this similarity is particularly remarkable when it is recalled that the cyanobacterial PII proteins interact with a set of receptors that are quite different to those of E. coli. This is consistent with the idea that recognition occurs through ¯exible loops and that the ¯exibility allows the recognition of multiple receptors (Xu et al., 1998). The T-loop in the present structures is disordered in one case and in the other is dependent upon crystal packing. Clearly, this piece of peptide is highly ¯exible. The C-terminal peptide is also capable of assuming multiple conformations although it tends to favour a 3 10 -helical or a strand structure. As noted earlier (Xu et al., 2001), it may favour a helical structure in solution, but this structure cannot be particularly stable. Indeed, the ¯exibility of this peptide may also be important for receptor recognition. The function of ATP in the PII-like proteins is not exactly clear, yet its binding pocket is highly conserved in terms of sequence (Xu et al., 1998). The cyanobacterial PII structures again show a highly conserved binding cleft. There are features of the cyanobacteria proteins that appear unique. A BLAST search to ®nd other proteins that were similar to Synechococcus PII gave a total of 167 hits. Amongst these sequences, cyanobacteria (Synechocystis, Nostoc punctiforme,Fremyella diplosiphon,Thermosynechococcus elongatus BP-1 and Prochlorococcus marinus) have a conserved Gln at the position 57. All proteobacteria and actinobacteria have a proline at position 57 in their GlnB and GlnK proteins. Other organisms have a variable nonproline at the corresponding position. These organisms include the archaebacteria, aqui®caceae, thermatogales, deinococci, red algae and dicotyledenous plants. Like the cyanobacteria proteins, these proteins are not uridylylated as evidenced by the fact that glnD (coding for UTase) does not exist in their genomes. Residue 57 is located in a section of the molecule that is likely to effect receptor recognition. We suggest that residue 57 may be involved in recognition events that are speci®c for cyanobacteria. We thank the Australian Research Council for supporting this work. We thank Paul Patch for technical assistance. We also thank the High Performance Computing Section of James Cook University for support and the Supercomputing Facility at the Australian National University for a grant of time on their machines. References Aldehni, M. F., Sauer, J., Spielhaupter, C., Schmid, R. & Forchhammer, K. (2003). J. Bacteriol. 185, 2582±2591. Acta Cryst. (2003). D59, 2183±2190 Xu et al. PII proteins 2189 research papers Figure 6 Ni 2+ ion linking the N- and C-termini in SnPII. (a) Location of the metalbinding site. (b) Close-up showing the metal coordination. Unbonded red spheres represent solvent molecules. research papers 2190 Xu et al. PII proteins Acta Cryst. (2003). D59, 2183±2190 Arcondeguy, T., Jack, R. & Merrick, M. (2001). Microbiol. Mol. Biol. Rev. 65, 80±105. Atkinson, M. R. & Ninfa, A. J. (1999). Mol. Microbiol. 32, 301±313. Benelli, E. M., Buck, M., Polikarpov, I., de Souza, E. M., Cruz, L. M. & Pedrosa, F. O. (2002). Eur. J. Biochem. 269, 3296±3303. Bru Ènger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W., Jiang, J.-S., Kuszewski, J., Nilges, N., Pannu, N. S., Read, R. J., Rice, L. M., Simonson, T. & Warren, G. L. (1998). Acta Cryst. D54, 905±921. Carr, P. D., Cheah, E., Suffolk, P. M., Vasudevan, S. G., Dixon, N. E. & Ollis, D. L. (1996). Acta Cryst. D52, 93±104. Cheah, E, Carr, P. D., Suffolk, P. M., Vasudevan, S. G., Dixon, N. E. & Ollis, D. L (1994). Structure,2, 981±990. Collaborative Computational Project, Number 4 (1994). Acta Cryst. D50, 760±763. Cowtan, K (1994). Jnt CCP4/ESF±EACBM Newsl. Protein Crystallogr. 31, 34±38. Forchhammer, K. (2003). Symbiosis,35, 101±115. Forchhammer, K. & de Marsac, N. T. (1995a). J. Bacteriol. 177, 5812± 5817. Forchhammer, K. & de Marsac, N. T. (1995b). J. Bacteriol. 177, 2033± 2040. Garcia-Dominguez, M. & Florencio, F. J. (1997). Plant Mol. Biol. 35, 723±734. Heeswijk, W. C. van, Stegeman, B., Hoving, S., Molenaar, D., Kahn, D. & Westerhoff, H. V. (1995). FEMS Microbiol. Lett. 132, 153±157. Hisbergues, M., Jeanjean, R., Joset, F., de Marsac, N. T. & Be Âdu, S. (1999). FEBS Lett. 463, 216±220. Hooft, R. W. W., Vriend, G., Sander, C. & Abola, E. E. (1996). Nature (London),381, 272. Irmler, A. & Forchhammer, K. (2001). Proc. Natl Acad. Sci. USA,98, 12978±12983. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991). Acta Cryst. A47, 110±119. Kabsch, W. (1976). Acta Cryst. A32, 922±923. Laskowski, R. A., MacArthur, M. W., Moss, D. W. & Thornton, J. M. (1993). J. Appl. Cryst. 26, 283±291. Lee, H. M., Flores, E., Forchhammer, K., Herrero, A. & de Marsac, N. T. (2000). Eur. J. Biochem. 267, 591±600. Lee, H. M., Flores, E., Herrero, A., Houmard, J. & de Marsac, N. T. (1998). FEBS Lett. 427, 291±295. Love, C. A., Lilley, P. E. & Dixon, N. E. (1996). Gene,176, 49±53. Matthews, B. W. (1968). J. Mol. Biol. 33, 491±497. Merrick, M. J. & Edwards, R. A. (1995). Microbiol. Rev. 59, 604±622. Navaza, J. (1994). Acta Cryst. A50, 157±163. Ninfa, A. J. & Atkinson, M. R. (2000). Trends Microbiol. 8, 172±179. Otwinowski, Z. & Minor, W. (1997). Methods Enzymol. 276, 307±326. Ruppert, U., Irmler, A., Kloft, N. & Forchhammer, K. (2002). Mol. Microbiol. 44, 855±864. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989). Molecular Cloning: A Laboratory Manual, 2nd ed. Cold Spring Harbor: Cold Spring Harbor Laboratory Press. Smith, C. S., Aalim, M. W. & Moorhead, G. B. G. (2003). Plant J. 33, 353±360. Vagin, A. & Teplyakov, A. (2000). Acta Cryst. D56, 1622±1624. Xu, Y., Carr, P. D., Huber, T., Vasudevan, S. G. & Ollis, D. L. (2001). Eur. J. Biochem. 268, 2028±2037. Xu, Y., Cheah, E., Carr, P. D., van Heeswijk, W. C., Westerhoff, H. V., Vasudevan, S. G. & Ollis, D. L. (1998). J. Mol. Biol. 282, 149± 165.