A two-component system mediates developmental regulation of biosynthesis of a heterocyst polysaccharide.

Some cyanobacteria couple oxygenic photosynthesis in vegetative cells with O2-sensitive N2 fixation in differentiated cells called heterocysts. Heterocyst differentiation involves extensive biochemical and structural changes that collectively permit heterocysts to assimilate N2 aerobically and supply the products of N2 fixation to vegetative cells. HepK and DevR are required for the development of functional heterocysts in Anabaena and Nostoc, respectively. We show that HepK is an autokinase and that Anabaena DevRA is its cognate response regulator, together comprising part or all of a two-component system that mediates developmental regulation of biosynthesis of a heterocyst envelope polysaccharide. Recombinant N-hexahistidine-tagged HepK (H6HepK), the cytoplasmic portion H6'HepK of H6HepK, H6DevR, and H6DevRA were overexpressed in Escherichia coli and purified to homogeneity. H6'HepK, but not H6HepK, autophosphorylates with [gamma-32P]ATP. ADP, specifically, elicits dephosphorylation of phosphorylated H6'HepK. The phosphoryl group of H6'HepK is transferred rapidly and efficiently to both H6DevR and H6DevRA but not to His-tagged OmpR, whose cognate sensor kinase is EnvZ. Sequence comparisons, the results of site-specific mutagenesis, and tests of chemical stability support identification of HepK-His348 and DevR-Asp53 as the phosphorylated residues. The mutation HepK-H348A abolishes both in vitro autokinase activity and in vivo functionality of HepK. Heterocysts of both hepK Anabaena and devRA Anabaena lack an envelope polysaccharide layer and are nonfunctional. Consistent with the normal site of deposition of that polysaccharide, a hepK::gfp transcriptional fusion is expressed principally in proheterocysts. HepK/DevRA is the first two-component system identified that regulates the biosynthesis of a polysaccharide as part of a patterned differentiation process.

When combined nitrogen becomes limiting, Anabaena species and their close relatives respond by initiating a developmental program that results in the production of terminally differentiated, nitrogen-fixing cells called heterocysts. Normally, approximately 10% of the cells become heterocysts that are present singly at semiregular intervals along the filaments, forming a spacing pattern. By sequestering nitrogenase within heterocysts, Anabaena spp. can carry out, simultaneously, oxygenic photosynthesis and the O 2 -labile assimilation of N 2 . Differences between heterocysts and their progenitor vegetative cells enable the heterocysts to provide a micro-oxic environment for nitrogenase (1). They lack the oxygen-producing reaction of photosystem II; respire rapidly; and have, outside of their cell wall, a layer of glycolipids that impedes the entry of oxygen and an outer, protective layer of polysaccharide (2,3).
Genes involved specifically in heterocyst differentiation and spacing have been cloned and characterized (4 -9). However, no overall regulatory strategy has been identified, and the detailed mechanisms that regulate differentiation remain largely unknown. It has been shown, however, that transcription of hepA, whose product is a member of the family of ABC transporters and is essential for biosynthesis of heterocyst envelope polysaccharide (4), is controlled by (i) HepK, a putative proteinhistidine kinase (10); (ii) HepC, which resembles galactosyl-PPundecaprenol synthetase (10); and (iii) two unusual site-specific DNA-binding proteins (11).
Polysaccharides, a major class of macromolecules, have great structural diversity. Once thought of only as reserve materials or inert structural elements, they are now recognized as playing an important role in the growth and development of plants (12,13), cell recognition and adhesion (14), hemostasis and lipid metabolism (15,16), pathogenesis (17), symbiosis (18,19), and defense responses (20,21). Different polysaccharides have been linked to carcinogenesis (15,22) and shown to have anticancer activity in vitro, in vivo, and in human clinical trials (23,24). Far more research on polysaccharides has emphasized their structure and biosynthesis than the regulation of their biosynthesis. Developing heterocysts provide a model of patterned, functional differentiation in which synthesis of a specific polysaccharide is an essential feature. The structure of heterocyst envelope polysaccharide has been determined for several species (25,26), the broad outlines of its biosynthesis have been examined (27), and a cluster of genes has been presumptively identified (28,29) that is involved in its biosynthesis. However, the regulation of its synthesis remains largely uncharted.
Two-component phosphorelay regulatory systems are the principal means for coordinating responses to environmental changes in bacteria and are found also in plants, fungi, protozoa, and amoebae (30 -32). Typically, such a system consists of a membrane-associated sensor kinase and its cognate response regulator. Most known sensor kinases consist of a cytoplasmic N terminus, two transmembrane segments linked by a periplasmic bridge, and a cytoplasmic, C-terminal, transmitter portion with a conserved histidine residue. Signal reception by the kinase stimulates an ATP-dependent autophosphorylation of that histidine. Transfer of the phosphate group to a con-served aspartyl residue of the response regulator renders the latter functional. Most known response regulators function as transcription factors whose effector domains bind DNA and regulate the expression of certain genes, effecting an adaptive response (33). However, some response regulators lack a known effector domain (32), and some effector domains are enzymatically active (34,35). When signaling subsides, both components undergo dephosphorylation, inactivating the response regulator.
Genomic sequence data have identified no presumptive vertebrate, nematode, or fruit fly genes that encode two-component phosphorelay proteins (36 -38). Because some two-component phosphorelay proteins are essential for the virulence of microbial pathogens including human fungal and bacterial pathogens, novel anti-microbial drugs targeted to two-component phosphorelay systems may prove highly specific for microbial pathogens. Although more than 1000 two-component phosphorelay proteins have been predicted from genomic data, very few have an identified function and have been shown experimentally to participate in phosphorelays. Some predicted homologs of histidine kinases may not be kinases or may not phosphorylate a histidine residue (32). For example, Azotobacter vinelandii NifL contains five conserved blocks of amino acids characteristic of histidine sensor kinases (39) but has not been observed to autophosphorylate (40,41), and DivL (42), pyruvate dehydrogenase kinase (43), and plant phytochromes (44) phosphorylate a tyrosine or serine rather than a histidine.
Predicted proteins HepK from Anabaena sp. strain PCC 7120 and DevR from Nostoc punctiforme strain ATCC 29133 resemble respectively a sensory protein-histidine kinase (10) and a response regulator that lacks known DNA-binding domains (45). The amino acid sequence of Anabaena sp. ortholog DevR A of DevR but for a possible 30-amino acid, N-terminal extension, is 93% identical and 98% similar to that of Nostoc DevR. Unlike most response regulators, DevR and DevR A have a 7-amino acid peptide, 57 SRSVYQG 63 , in the ␥-turn loop region. DevR, but not DevR⌬7, from which that peptide was deleted, is weakly phosphorylated in vitro by Escherichia coli sensor kinase EnvZ, a result that supported the interpretation of Hagen and Meeks (46) that DevR is a response regulator in an unidentified phosphorelay system that controls heterocyst maturation in N. punctiforme strain ATCC 29133. However, no cognate histidine kinase was identified for DevR. No gene that evidently encodes a response regulator neighbors hepK. devR A is 1.4 Mb distant from hepK (28). hepK Anabaena sp. and devR N. punctiforme cannot assimilate N 2 aerobically (10,45). We present evidence that Anabaena sp. HepK is a histidine kinase, that DevR A is its cognate response regulator, and that the two comprise part or all of a two-component system that controls the biosynthesis of heterocyst envelope polysaccharide.

EXPERIMENTAL PROCEDURES
Culture and Preparation for Electron Microscopy-Anabaena sp. strain PCC 7120 and derivatives of it were grown at 30°C in the light at approximately 140 microeinsteins m Ϫ2 s Ϫ1 (Li-Cor Quantum Radiometer/Photometer model LI-185A; Lincoln, NE) on AA ϩ N agar (47) in the presence of appropriate antibiotics or on a rotary shaker in medium AA/8 or AA/8 ϩ N (47) plus appropriate antibiotics. Wild-type and devR A Anabaena sp. strains and wild-type N. punctiforme strain ATCC 29133 and its devR derivative UCD311 were prepared for electron microscopy (48) after the following regimens of growth. Anabaena sp. strains grown 5 days in AA/8 ϩ N (plus, for the mutant strain, 2 g of spectinomycin sulfate ml Ϫ1 ) were sedimented, washed three times with AA/8, and incubated for 3 days on a Nuclepore Rec-85 membrane atop AA agar. Wild-type N. punctiforme was grown for 5 days in AA/8. UCD311 was incubated 2 days in AA/8 after 7 days of growth in AA/8 ϩ N plus 10 g of neomycin sulfate ml Ϫ1 and three washes with AA/8.
DNA Manipulation-Recombinant DNA procedures were performed in a standard manner (49). The enzymes were purchased from New England Biolabs, Invitrogen, and Roche Applied Science and used as recommended by the manufacturers.
Construction of Expression Plasmids-Truncated hepK (ЈhepK) encoding residues 267-575 of HepK that lack the two presumptive transmembrane regions and full-length hepK (GenBank TM accession number U68034) were amplified by PCR with, respectively, primer pairs 5Ј-G-GAATTCCATATGCGGACTAGTCGAGCGATCGCT-3Ј and 5Ј-GAAGA-TCTTAACTTTGCTCCTGAAGTGGCAGG-3Ј (CPW99), and 5Ј-GGAAT-TCCATATGCAAACTCAGCAGCCGATCCCTG-3Ј and CPW99 (introduced NdeI and BglII sites are underlined) with pRL2079 (10) as template. The PCR products were cloned between the NdeI and BamHI sites of plasmid pET-14b (Novagen, Inc., Madison, WI), which provided an N-terminal hexahistidine tag. To reduce the extent of PCR product that had to be validated by sequencing, the region from MfeI to BstXI in intact hepK was replaced by a portion of wild-type hepK as was a smaller portion in truncated hepK. The resulting plasmids were denoted pRL2406 and pRL2433, respectively. devR and devR⌬21 (encoding DevR⌬7) of N. punctiforme were PCR-amplified from plasmids pSCR169 and pSCR343 (Ref. 46; gifts of J. C. Meeks), respectively, as templates with the primers 5Ј-GGAATTCCATATGAAAACTGTTTTA-ATTGTC-3Ј and 5Ј-GAAGATCTTATGATTGGCCGTCTGTGG-3Ј and were cloned between the NdeI and BamHI sites of pET-14b, producing pRL2461 and pRL2487. pSCR351, encoding H 6 DevR-D53Q, was also from Dr. Meeks. devR A , PCR-amplified from total Anabaena sp. DNA with the primers 5Ј-GGAATTCCATATGACAGCAAAGCATATTGATA-CTA-3Ј and 5Ј-CGCGGATCCTCAGTGGTTGTCACTGGGGAGGAGAG-3Ј, was cloned between the NdeI and BamHI sites of pET-14b to produce pRL2749 for production of H 6 DevR A . The ompR gene (Gen-Bank TM accession number J01656) of E. coli strain MC4100 in plasmid pLAN801 (a gift of Dr. M. Igo) was amplified by PCR with the primers 5Ј-GGGAATTCCATATGCAAGAGAACTACAAGATTCTG-3Ј and 5Ј-C-CCAAGCTTTCATGCTTTAGAGCCCTCCGG-3Ј. The PCR product was first cloned into the EcoRV site of pBluescript SK ϩ (Stratagene, La Jolla, CA) to produce pRL2483 for DNA sequencing; then an NdeI-BamHI fragment containing ompR was transferred between the NdeI and BamHI sites of pET-14b, yielding pRL2484. All of the above PCR products were proven error-free by DNA sequencing.
Site-directed Mutation of hepK-Primer-mediated PCR mutagenesis (QuikChange site-directed mutagenesis; Stratagene) was used to change the conserved His 348 to Ala at the presumptive site of phosphorylation of HepK to produce a mutant truncated HepK protein, which we denote ЈHepK-H348A. The mutagenic oligonucleotides 5Ј-ATTTCT-TGCCAACGTTAGTGCTGAGTTGCGTAC-3Ј and 5Ј-GTACGCAACT-CAGCACTAACGTTGGCAAGAAAT-3Ј (codon 348, CAT 3 GCT) with pRL2406 as template were used for PCR; the product was cloned in pET-14b, producing pRL2454. The sequence of the 310-bp BstEII-BstXI fragment of pRL2454 was shown to be identical to the corresponding fragment of pRL2406, except for the H348A substitution, by DNA sequencing. Transfer of the BstEII-BstXI fragment from pRL2454 to pRL2406 produced pRL2457, from which H 6 ЈHepK-H348A was produced and purified as had been H 6 ЈHepK.
Test of Complementation of hepK Mutant Y7 by hepK-H348A-Transposon-generated hepK mutant Y7 cannot grow aerobically on N 2 , but Y7 (pRL2078) can (10). pRL2480, identical to pRL2078 but for an H348A change in hepK, was constructed by transferring the 4.9-kb, hepKbearing BamHI DNA fragment of pRL2078 to pRL57 (50), which lacks BstEII and BstXI sites; replacing the 310-bp BstEII-BstXI fragment of hepK by the corresponding fragment of pRL2457; and then replacing the 4.9-kb BamHI fragment of pRL2078 with the resulting BamHI fragment. pRL2480 was transferred by conjugation (51) into Y7 with selection on AA plus nitrate agar plates supplemented with 30 g of neomycin sulfate ml Ϫ1 to select for Y7 and 10 g of spectinomycin sulfate ml Ϫ1 to select for pRL2480. 10 days later, 10 resulting colonies were restreaked onto the same medium for proliferation and were then spotted onto AA agar plates without or with nitrate (the latter was supplemented with neomycin and spectinomycin as before).
Assays of Phosphorylation and Transphosphorylation-Unless otherwise specified, phosphorylation reaction mixtures contained 25 mM Tris-HCl, pH 8.0, 50 mM NaCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.5 mM EDTA, and 30 -80 M [␥-32 P]ATP (6,000 Ci mmol Ϫ1 ). H 6 HepK, H 6 DevR, or derivatives of them, H 6 DevR A , H 6 EnvZc, and H 6 OmpR were added to final concentrations of 2.5 M or as indicated. The reactions were initiated by the addition of the radioactive substrate, incubated at room temperature (approximately 24°C) for different times, and stopped by the addition of an equal volume of 2ϫ concentrated SDS-PAGE loading buffer (50 mM Tris-HCl, pH 6.8, 4% SDS, 10% glycerol, 200 mM dithiothreitol, 0.003% bromphenol blue). The reactions, once stopped, were immediately subjected to SDS-PAGE. After completion of electrophoresis, the gels were electroblotted onto polyvinylidene difluoride membranes (Bio-Rad) at 200 mA in protein-transfer buffer (25 mM Tris, 250 mM glycine, 5% methanol, 0.1% SDS, pH 8.3) at 4°C for 1 h, and the membrane was then air-dried at room temperature. The radioactivity of proteins attached to polyvinylidene difluoride membranes was determined qualitatively by autoradiography with Hyperfilm (Amersham Biosciences) or quantitatively with a PhosphorImager model V5.6 (Molecular Dynamics, Sunnyvale, CA). To show the protein bands, the polyvinylidene difluoride membrane was later stained with Coomassie Brilliant Blue R-250 for 2 min and then destained with 50% methanol. In time course experiments, portions of a sample were removed, mixed with loading buffer, and kept on ice until all of the samples were taken.
To compare transfer of 32 P from H 6 ЈHepK-32 P to H 6 DevR and to H 6 DevR⌬7, one 13-l fraction and two 40-l fractions were removed from a solution of 9 g of phosphorylated H 6 ЈHepK in 100 l of standard phosphorylation buffer. Six g of H 6 DevR or H 6 DevR⌬7 in 12 l of standard phosphorylation buffer were added to separate 40-l fractions of phosphorylated H 6 ЈHepK. At the indicated reaction times, 15-l portions of samples were rapidly transferred to an equal volume of 2ϫ concentrated SDS-PAGE loading buffer. The mixtures were kept on ice until all of the portions had been removed and were then subjected to SDS-PAGE.
Stability of Phosphate Linkages in Phosphorylated H 6 ЈHepK, H 6 DevR A , and H 6 DevR, and Effects of Nucleoside Diphosphates on Dephosphorylation of H 6 ЈHepK-32 P-To examine the chemical stability of their phosphate linkages, 30 pmol of H 6 ЈHepK or 50 pmol of H 6 DevR A or H 6 DevR with 20 pmol of H 6 ЈHepK were phosphorylated for 30 min in 30 l of reaction mixture. The reactions were terminated by adding an equal volume of 2ϫ concentrated SDS-PAGE loading buffer. Samples of 10 l were pipetted into microcentrifuge tubes containing 2 l of 1 N HCl, 0.15 N HCl, water, or 3 N NaOH, giving final pH values of 0.9, 2.7, 7.2, or 12.5, respectively. pH values were confirmed with Color pHast indicator strips (EM Science, Darmstadt, Germany). The mixtures were incubated for 60 min in a water bath at 42°C. Another sample, to which 2 l of water was added, was kept on ice. The samples were separated by SDS-PAGE and analyzed for radioactivity.
The kinetics of loss of 32 P from H 6 ЈHepK-32 P and H 6 DevR-32 P at pH 12.5 were assessed as follows. H 6 ЈHepK (35 pmol) was phosphorylated for 30 min in a 30-l reaction mixture; in some experiments the 35 pmol of H 6 ЈHepK-32 P in 20 l was then used to phosphorylate 35 pmol of H 6 DevR for 30 min in a 30-l reaction mixture, and the protein or proteins were then denatured at neutral pH by the addition of 30 l of 2ϫ concentrated loading buffer (pH 6.8). Each resulting 60-l mixture, prewarmed for 2 min at 42°C, was adjusted to pH 12.5 by addition of 12 l of 3 N NaOH. pH values were confirmed with Color pHast indicator strips. At intervals thereafter (see Fig. 6), an 8-l sample was withdrawn and introduced into a microcentrifuge tube containing 1.33 l of 3 N HCl to readjust the pH to 7.0. The samples were immediately placed on ice until sampling was completed. All of the samples were separated by SDS-PAGE and analyzed for radioactivity.
ADP, CDP, GDP, or TDP (sodium salts; Sigma; sodium ADP was also purchased from TCI America for confirmation) at a final concentration of 0.5 mM was incubated for 5 min in a 15-l phosphorylation reaction mixture containing 2 g of H 6 ЈHepK-32 P. The reactions were stopped by addition of 2ϫ concentrated SDS-PAGE loading buffer. The samples subjected to SDS-PAGE were analyzed for radioactivity.
Inactivation of devR A -A 1415-bp DNA fragment containing devR A and DNA downstream from it was amplified by PCR from genomic DNA of Anabaena sp. with the primers 5Ј-TAAATAAAATTCTCATGAAAA-CTG-3Ј and 5Ј-GAGCAAGACTGGCTAGCCAGTT-3Ј (BspHI site underlined). The PCR product was first cloned into the EcoRV site of pBluescript SK ϩ to produce pRL2478 for DNA sequencing. Then a BspHI-BamHI fragment containing the PCR product, which was free of errors, was transferred between the BspHI and BglII sites of pRL278 (70) to produce pRL2479. devR A was inactivated by insertion of a 2.7-kb, SmaI-gfp-omega-PvuII cassette from pRL2379 (8) into the unique NruI site of pRL2479, 110 bp 5Ј from the stop codon of devR A . The resulting plasmid, denoted pRL2481a, was transferred to wild-type Anabaena sp. by conjugation, and a double recombinant (devR A ) strain denoted DR2481a was isolated as described (50). The structure of the recombinant was confirmed by diagnostic PCR using primers 5Ј-CGGGATC-CCGTGGAGAGAAGGTTTAA-3Ј and 5Ј-CGGGATCCTCAGTGGTTGT-CACTGGGGAG-3Ј and, as templates, total DNA isolated from the wild-type strain and DR2481a.
Construction of hepK::gfp and Fluorescence Microscopy-A 4.9-kb BamHI fragment of pRL2078, containing the 1.7-kb hepK-encoding region and upstream sequence, was cloned into the BamHI site of shuttle vector pRL57 (50) to produce pRL2419. The gfp-omega cassette from pRL2379 (8) was inserted into the blunted, unique BstXI site 1.2 kb downstream from the start codon of hepK in pRL2419. The resulting plasmid, denoted pRL2437, was transferred into wild-type Anabaena sp. strain PCC 7120 by conjugation (51). Fluorescence microscopy and photography of the hepK::gfp strain was performed as described (53).
Stability of Phosphate Linkages to H 6 ЈHepK, H 6 DevR A , and H 6 DevR-The chemical stability of the phosphoryl bonds in phospho-H 6 ЈHepK, -H 6 DevR A , and -H 6 DevR was tested. Phosphorylated amino acid residues in proteins are commonly classified into O-phosphates, phosphomonesters formed by phos-phorylation of amino acids Ser, Thr, and Tyr; N-phosphates, phosphoramidates, produced by phosphorylation of Arg, His, and Lys; and acyl phosphates, phosphate anhydrides generated by phosphorylation of Asp and Glu. Phospho-His and -Lys are stable under alkaline conditions but sensitive to mildly acidic conditions, whereas the opposite is true for phospho-Asp and -Glu (which are, however, unstable at pH 0.9); and phospho-Arg is unstable under both conditions (55,56). H 6 ЈHepK-32 P and H 6 DevR-32 P were both dephosphorylated within 1 h at the highly acidic pH, 0.9 (Fig. 6, A and B, lanes 2). H 6 ЈHepK-32 P lost more than ϳ90% of its phosphate within 1 h at pH 2.7 (Fig.  6A, lane 3) but was relatively stable at pH 12.5 (Fig. 6A, lane 5). H 6 DevR-32 P (Fig. 6B, lane 3) and H 6 DevR A -32 P (Fig. 6C, lane 2) were relatively stable at pH 2.7 but unstable at pH 12.5 (Fig. 6,  B, lane 5, and C, lane 4). H 6 DevR A -32 P (data not shown), like H 6 DevR-32 P (Fig. 7B), showed no residual radioactivity after 2 min at 42°C and pH 12.5. H 6 ЈHepK-32 P did not lose significantly more radioactivity during 2-60 min at pH 12.5 and 42°C (Fig. 7A, lanes 3-7) than at pH 7.0 and either 42 or 0°C (Fig. 7A, lanes 1 and 2)   DevR and of DevR A are phosphorylated. This inference was tested by use of H 6 DevR-D53Q (46) and construction and use of H 6 ЈHepK-H348A (the numbers 348 and 53 refer to the untruncated amino acid sequences in the absence of the hexahistidine tag). Unlike H 6 ЈHepK (Fig. 8, A and B, lane 1), H 6 ЈHepK-H348A overexpressed in E. coli and purified to homogeneity (Fig. 8B, lanes 3 and 4) failed to autophosphorylate (Fig. 8A,  lane 3). H 6 DevR-D53Q, similarly overexpressed and purified to homogeneity (data not shown), unlike H 6 DevR, did not accept 32 P (Fig. 8C, lanes 2 and 3). hepK mutant Y7 of Anabaena sp. is neomycin-resistant and unable to assimilate N 2 aerobically (10) and is spectinomycin-sensitive and therefore unable to grow in the presence of neomycin, spectinomycin, and fixed nitrogen (Fig. 9C, circle). Plasmid pRL2078, which bears wildtype hepK (10) and confers resistance to spectinomycin, complemented Y7, as shown by the dark green growth of Y7 (pRL2078) (Fig. 9D, circle) in the absence of fixed nitrogen. Plasmid pRL2480 is identical to pRL2078 except that it en- FIG. 9. hepK-H348A fails to complement hepK mutant Y7. Ten independent transformants of mutant Y7 bearing pRL2480 selected and grown in the presence of nitrate were inoculated as pale, chlorophyll-containing spots to (A, day 0) AA medium with nitrate plus agar supplemented with 30 g of neomycin sulfate ml Ϫ1 and 10 g of spectinomycin sulfate ml Ϫ1 and (B, day 0) AA plus agar only. The medium shown in A then received a single spot (circle in C) of neomycinresistant, spectinomycin-sensitive mutant Y7, and the medium shown in B received a spot (circle in D) of Y7 bearing pRL2078, which encodes HepK (10). pRL2480 is identical to pRL2078 except that it encodes a HepK-H348A variant of HepK. The pRL2480-transformed Y7 that had been fed nitrate grew densely (C, day 5), whereas neomycin-resistant, spectinomycin-sensitive mutant Y7 (circle) failed to grow, illustrating that pRL2480 had conferred resistance to spectinomycin. In contrast, transformants of Y7 with pRL2480 had yellowed and had grown little if at all without nitrate, whereas Y7 with pRL2078 (circle) grew well (D, day 5). codes a HepK-H348A variant of HepK. In contrast to pRL2078, pRL2480 failed to complement Y7, as shown by the lack of growth in the absence of fixed nitrogen (Fig. 9D) of 10 randomly chosen derivatives of Y7 that bear pRL2480 and are therefore able to grow in the presence of fixed nitrogen, neomycin, and spectinomycin (Fig. 9C).
ADP Elicits Dephosphorylation of H 6 ЈHepK-32 P but Has No Evident Effect on Phosphotransfer from H 6 ЈHepK-32 P to DevR-Because autophosphorylation of ЈHepK presumably produces ADP, we tested whether ADP affects autophosphorylation of, or phosphate transfer from, H 6 ЈHepK-32 P. Unlike incubation with CDP, GDP, or TDP (Fig. 10A, lanes 1 and 3-5), incubation of a solution of phosphorylated H 6 ЈHepK for 5 min with 0.5 mM ADP resulted consistently in a loss of ϳ90% of the 32 P from H 6 ЈHepK-32 P (Fig. 10A, lanes 1 and 2). The loss was even greater after 0.5 h (Fig. 10B, lanes 3 and 4), whereas ADP negligibly affected net phosphate transfer from H 6 ЈHepK-32 P to H 6 DevR (Fig. 10B, lanes 1 and 2). In the absence of H 6 ЈHepK Inactivation of devR A , devR, or hepK Blocks Formation of the Heterocyst Envelope Polysaccharide Layer-Heterocysts of hepK mutant Y7 of Anabaena sp. form a laminated layer of glycolipids but no polysaccharide layer and fail to fix N 2 aerobically (10). Inactivation of devR A by double reciprocal recombination with plasmid pRL2481a (see "Materials and Methods") was confirmed by diagnostic PCR (data not shown). devR A Anabaena sp. also cannot grow aerobically on N 2 . As in hepK Anabaena sp. (10), heterocyst envelopes of devR A Anabaena sp. (Fig. 11, C and D) and devR N. punctiforme (Fig. 11, G and I) retain a laminated glycolipid layer but lack a polysaccharide layer (Fig. 11, compare A, B, E, and F, respectively).
In Vivo Localization of Expression of hepK-Plasmid pRL2437 bears a hepK::gfp transcriptional fusion (gfp encodes green fluorescent protein) (57). Fluorescence from that transcriptional fusion was observed primarily in single proheterocysts (Fig. 12D, arrow P) but sometimes in paired cells (Fig. 12, B and C, arrow P), of which at least one was a proheterocyst. Slight fluorescence, of questionable significance in this highly autofluorescent organism, originated from vegetative cells (Fig.  12, A-D) and mature heterocysts (Fig. 12B, arrow H). DISCUSSION HepK resembles protein-histidine kinases of two-component regulatory systems (10), and DevR resembles certain response regulators (45). On the basis of the following observations, we conclude that HepK is a histidine kinase and that it and DevR A function in Anabaena sp. strain PCC 7120 as interacting components of a two-component regulatory system. First, H 6 ЈHepK autophosphorylates (Figs. 2-8 and 10) and serves as phosphodonor, rapidly and efficiently, to H 6 DevR A (Fig. 3, A and B) and H 6 DevR (its N. punctiforme ortholog; Fig. 3, C and D) but (unlike H 6 EnvZc) not to H 6 OmpR (Fig. 4A). Second, replacement of conserved amino acid His 348 of HepK by Ala prevents both in vitro phosphorylation (Fig. 8) and in vivo function (Fig.  9), and replacement of conserved Asp 53 of H 6 DevR by Gln prevents its activity as a phospho-acceptor from H 6 ЈHepK (Fig.  8) and its in vivo function in N. punctiforme (46). Third, like inactivation of hepK (10), inactivation of devR A or devR blocks formation of a heterocyst envelope polysaccharide layer, while abundant synthesis of the laminated structure composed of glycolipids continues (Fig. 11). Because DevR, the phenotype of a devR mutant, and the N. punctiforme HepK ortholog are highly similar to DevR A , the phenotype of devR A Anabaena, and HepK, respectively, it seems highly likely that DevR and the N. punctiforme HepK ortholog function and interact similarly to DevR A and HepK. We know of no other two-component system that regulates the biosynthesis of a polysaccharide as part of a patterned differentiation process.
Protein-histidine kinases VirA, ArcB, and EnvZ autophosphorylate in vitro when truncated by removal of transmembrane regions but not when intact (54,55,58). Similarly, we observed phosphorylation in vitro by truncated but not by intact H 6 HepK (Fig. 2). Perhaps intact HepK cannot fold properly in vitro; when overexpressed in E. coli, H 6 HepK formed mostly inclusion bodies, whereas H 6 ЈHepK did not. Alternatively, intact HepK may autophosphorylate only in response to a signal specific to heterocyst development that is lacking in vitro. Sequence alignments (10,45) suggested that His 348 of HepK and Asp 53 of DevR A and DevR may be phosphorylated. Concordantly, linkage of 32 P to H 6 ЈHepK is unstable at pH 2.7 and stable at pH 12.5 (Figs. 6 and 7), behavior typical of phosphorylated histidine or lysine residues, whereas H 6 DevR A -32 P and H 6 DevR-32 P are moderately stable at pH 2.7 and unstable at pH 12.5 (Figs. 6 and 7 and data not shown), as is characteristic of acyl phosphates (55,56). Dephosphorylation of both histidine N-32 P (as observed for H 6 ЈHepK-32 P; Fig. 6A, lane 2) and aspartate acyl-32 P (as observed for H 6 DevR-32 P; Fig. 6B, lane 2) at pH 0.9 is expected (55).
Response regulators such as DevR A and DevR that lack a known C-terminal effector domain may (i) interact with a downstream target only in response to phosphorylation (59) or (ii) act as an intermediate in a multicomponent phosphorelay system (33,60). Both DevR-D53Q (glutamine is uncharged and cannot be phosphorylated) and DevR-D53E (glutamate may structurally mimic phosphorylated aspartate) lack DevR function in vivo in N. punctiforme (46), suggesting that the function of DevR may depend on a role of DevR-Asp 53 as a phosphotransferase intermediate. That is, DevR, like Spo0F, may phosphorylate a hitherto unidentified DNA-binding protein (46).
The genome of Anabaena sp. strain PCC 7120 contains 73 genes for putative simple His kinases, 53 more for putative hybrid His kinases, and 77 genes for putative simple response regulators, in addition to other kinds of protein kinases (61). Few of these have a known function, and interacting pairs have not been demonstrated. Specificity of interaction of a particular regulator with a particular sensor in response to a given environmental stimulus may often be essential to avoid inappropriate responses, although certain response regulators may have to receive input signals from more than one sensor. Although there is very weak cross-talk in vitro between the noncognate histidine kinase EnvZ, whose cognate response regulator is OmpR (54), and H 6 DevR, no cross-talk was observed with H 6 DevR⌬7 (46). 1 Whereas the 7-amino acid peptide, SRS-VYQG, located in DevR and DevR A is absent from most known response regulators, a similar insertion is present in FruA, a putative response regulator in Myxococcus xanthus (62). Our biochemical data (Fig. 5) showed that deletion of that peptide from the ␥-turn loop region of H 6 DevR greatly decreased the rate and extent of phosphotransfer from H 6 ЈHepK to H 6 DevR⌬7 in vitro and so may influence the rate, and perhaps the specificity, of phosphotransfer from HepK in vivo. The lack of cross-talk between HepK and OmpR (Fig. 4A) confirmed the specificity of recognition between HepK and DevR A .
Although the presence of ADP results in dephosphorylation 1 R. Zhou and C. P. Wolk, unpublished data. of H 6 ЈHepK-32 P (Fig. 10), we saw no diminution of H 6 ЈHepKdependent phosphorylation of H 6 DevR (Fig. 10B, lanes 1 and  2), perhaps because the latter reaction takes place much more rapidly. On the basis of the observed effect of ADP in vitro, we conjecture that the ratio of ATP to ADP in vivo may regulate the level of phosphorylated H 6 ЈHepK.
Because a chromosomal hepK::gfp fusion was expressed too weakly for us to localize the resulting fluorescence, we expressed hepK::gfp in a pDU1-based plasmid. Unlike expression of a glnA fusion in a pDU1-based plasmid, which was localized in all cells of N 2 -fixing filaments (63) as expected on the basis of biochemical data (64,65), hepK was expressed primarily in proheterocysts (Fig. 12). This result appears consistent with the fact that it is in developing heterocysts that synthesis of heterocyst envelope polysaccharide, a process that is regulated by HepK (10), takes place.
As noted above, no human, nematode, or fruit fly sequences that presumptively encode two-component phosphorelay proteins have been identified. Because two-component phosphorelay systems are specifically involved with the establishment of virulence in some bacteria and fungi, sometimes via regulation of polysaccharides (66,67), drugs targeted to such systems may prove highly specific for microbial pathogens (68,69). Therefore, study of phosphorelay regulation of polysaccharide synthesis in Anabaena, a nonpathogen, may have relevance for drug development.