Bioinformatics and Experimental Analysis of Proteins of ...Bioinformatics and Experimental Analysis of Proteins of ... essential for bacteria in order to sense and respond to changes

JOURNAL OF BACTERIOLOGY, Jan. 2008, p. 613–624 Vol. 190, No. 20021-9193/08/$08.00�0 doi:10.1128/JB.01502-07Copyright © 2008, American Society for Microbiology. All Rights Reserved.

Bioinformatics and Experimental Analysis of Proteins ofTwo-Component Systems in Myxococcus xanthus�†Xingqi Shi, Sigrun Wegener-Feldbrugge, Stuart Huntley, Nils Hamann,

Reiner Hedderich, and Lotte Søgaard-Andersen*Department of Ecophysiology, Max Planck Institute for Terrestrial Microbiology, Karl-von-Frisch Str., 35043 Marburg, Germany

Received 18 September 2007/Accepted 2 November 2007

Proteins of two-component systems (TCS) have essential functions in the sensing of external and self-generated signals in bacteria and in the generation of appropriate output responses. Accordingly, in Myxo-coccus xanthus, TCS are important for normal motility and fruiting body formation and sporulation. Here weanalyzed the M. xanthus genome for the presence and genetic organization of genes encoding TCS. Two hundredseventy-two TCS genes were identified, 251 of which are not part of che gene clusters. We report that the TCSgenes are unusually organized, with 55% being orphan and 16% in complex gene clusters whereas only 29%display the standard paired gene organization. Hybrid histidine protein kinases and histidine protein kinasespredicted to be localized to the cytoplasm are overrepresented among proteins encoded by orphan genes or incomplex gene clusters. Similarly, response regulators without output domains are overrepresented amongproteins encoded by orphan genes or in complex gene clusters. The most frequently occurring output domainsin response regulators are involved in DNA binding and cyclic-di-GMP metabolism. Our analyses suggest thatTCS encoded by orphan genes and complex gene clusters are functionally distinct from TCS encoded by pairedgenes and that the connectivity of the pathways made up of TCS encoded by orphan genes and complex geneclusters is different from that of pathways involving TCS encoded by paired genes. Experimentally, we observedthat orphan TCS genes are overrepresented among genes that display altered transcription during fruitingbody formation. The systematic analysis of the 25 orphan genes encoding histidine protein kinases that aretranscriptionally up-regulated during development showed that 2 such genes are likely essential for viabilityand identified 7 histidine protein kinases, including 4 not previously characterized that have importantfunction in fruiting body formation or spore germination.

A fundamental property of cells is their ability to sense andrespond to external stimuli and self-generated signals. In thecase of bacteria, this ability maximizes their chances of survival.Signal transduction proteins have essential functions in stimu-lus sensing, information processing, and the generation of out-put responses. Despite the multitude of cues that bacteria needto monitor, the signal transduction schemes involved center ona few types (11, 65): ligand-regulated one-component systems,which consist of single protein molecules containing both asensing domain and an output domain; cyclic-di-GMP syn-thetases, phosphodiesterases, adenylate and guanylate cyclases,which act by modifying the level of secondary messenger mol-ecules; methyl-accepting chemotaxis proteins that modulatethe activity of chemosensory systems; and systems in whichinformation transfer depends on covalent modification byphosphorylation/dephosphorylation by means of either Ser/Thr/Tyr protein kinases or histidine protein kinases (HPKs).HPKs are the more common type of protein kinase in bacteriaand function as part of two-component systems (TCS) (11).Despite the fact that one-component systems are the numeri-cally dominant sensing systems in bacteria (65), TCS are

essential for bacteria in order to sense and respond to changesin the environment (58).

A typical TCS consists of an HPK and a cognate responseregulator (RR) which are encoded in the same operon andarchitecturally organized in a simple linear 1:1 phosphotrans-fer signal transduction pathway (58). HPKs are multidomainproteins and generally contain a nonconserved sensor domain,which is responsible for detecting a particular stimulus, and ahighly conserved kinase domain, which can be further subdi-vided into the HisKA domain, which is involved in dimeriza-tion and contains the conserved phosphorylatable histidineresidue, and the HATPase_c domain. Typical RRs are eithersingle-domain proteins consisting only of the conserved re-ceiver domain, which contains the conserved aspartate residuethat accepts the phosphoryl group from the histidine residue inthe cognate HPK, or multidomain proteins consisting of areceiver domain and a variable output domain. HPKs auto-phosphorylate in a stimulus-dependent manner on the con-served histidine residue in the HisKA domain using ATP as aphosphodonor. Subsequently, the phosphoryl group is trans-ferred to the conserved aspartate residue in the receiver do-main of the cognate RR, resulting in activation of the RR andthe generation of an appropriate output response. The phos-phorelay systems constitute a structurally more complex typeof TCS, with phosphotransfer occurring in three sequentialsteps (3). Generally, the architecture of these systems is similarto that of typical TCS and involves a linear phosphotransferscheme. Specifically, the phosphoryl group is first transferredfrom the conserved histidine in the HPK to the conserved

* Corresponding author. Mailing address: Department of Ecophysi-ology, Max Planck Institute for Terrestrial Microbiology, Karl-von-Frisch Str., 35043 Marburg, Germany. Phone: 49-6421-178201. Fax:49-6421-178209. E-mail: [email protected].

† Supplemental material for this article may be found at http://jb.asm.org/.

� Published ahead of print on 9 November 2007.

613

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

aspartate in a receiver domain, next to a conserved histidine ina phosphotransferase domain (Hpt), and finally to the con-served aspartate in a second RR. The domain organization ofproteins in phosphorelays is highly modular. Thus, all fourdomains involved in phosphotransfer may reside in separateproteins (5), the kinase and first receiver may be present in thesame protein (42), or the kinase, the first receiver, and the Hptdomain may be present in the same protein (64). It has beenargued that the specific advantage of phosphorelays over typ-ical TCS is that they allow the integration of several sensoryinputs in one signal transduction pathway (9). In TCS, as wellas in phosphorelays, the output response is determined by thephosphorylation status of the final RR and typically involvesalterations in gene expression, in protein-protein interactions,or in enzyme activity. In addition to the linear 1:1 TCS and1:1:1:1 phosphorelays, TCS and phosphorelays may also beorganized as branched pathways. In the one-to-many pathways(54), one HPK has several cognate RRs, as exemplified byCheA (58) and ArcB (41) in Escherichia coli. This pathwayconnectivity would be suited for generating several responsesto one input signal. In the many-to-one pathway (54), one RRhas several cognate HPKs, as exemplified by Spo0F in Bacillussubtilis (5) and DosR in Mycobacterium tuberculosis (31). Thispathway connectivity would be suited for integrating severalinput signals to generate one output response.

Here we have focused on an analysis of TCS in the gram-negative deltaproteobacterium Myxococcus xanthus. M. xan-thus has a highly complex life cycle that involves social behavior(23). In the presence of nutrients, cells grow and divide, and ifpresent on a solid surface, they form coordinately spreadingcolonies. In response to starvation, a developmental program isinitiated that culminates in the formation of multicellular,spore-filled fruiting bodies. This developmental program in-volves two temporally and spatially coordinated morphoge-netic events, aggregation of cells into fruiting bodies andsporulation of cells that have accumulated inside fruiting bod-ies. Cells that remain outside the fruiting bodies do not sporu-late (43). Fruiting body formation is initiated by the RelA-dependent accumulation of the intracellular signalingmolecule(s) guanosine penta- and tetraphosphate [(p)ppGpp](16, 53). In addition, two intercellular signals, the A- andC-signals, are important for fruiting body formation. The A-signal consists of a mixture of amino acids and peptides (32)and is part of a system that monitors the density of starvingcells (32, 33). The C-signal is a 17-kDa protein (27, 38) thatinduces and coordinates aggregation and sporulation (26, 30,37). Fruiting body formation is accompanied by changes ingene expression, and several of the genes required for fruitingbody formation are transcriptionally regulated in response tostarvation (19, 29). The formation of spreading colonies, aswell as fruiting body formation, depends on the functionality ofthe A- and S-gliding motility systems (18).

Forward- and reverse-genetics approaches have identified 35TCS genes that are important for fruiting body formation (seeTables S3, S4, and S5 in the supplemental material for a list ofthese genes). These effects on fruiting body formation, in sev-eral cases, e.g., the RRs DigR (44) and AglZ (67), are likely tobe indirect and caused by an effect on gliding motility. How-ever, most other TCS mutants display only developmental de-fects, suggesting that the corresponding proteins are directly

involved in development. More than half of the TCS genesrequired for fruiting body formation are encoded by orphanTCS genes or by TCS genes in complex gene clusters (SeeMaterials and Methods for a definition of TCS genes), and thearchitecture of the corresponding signal transduction pathwaysremains poorly understood.

To begin to address the function of the remaining TCS genesin M. xanthus, we used bioinformatics in combination withfunctional analyses. We report that 71% of the TCS genes areorganized in unusual manners as orphan genes or in complexgene clusters whereas the remaining 29% display the standardpaired gene organization. Our bioinformatics analyses suggestthat TCS proteins encoded by orphan genes and complex geneclusters are functionally distinct from TCS proteins encoded bypaired genes and that the connectivity of the pathways madeup of TCS proteins encoded by orphan genes and complexgene clusters is different from that of pathways involving TCSproteins encoded by paired genes. Experimentally, we foundthat orphan TCS genes are overrepresented among TCS genesthat display altered transcription during fruiting body forma-tion. The systematic analysis of 25 orphan HPKs that are tran-scriptionally up-regulated during development led to the iden-tification of 2 HPKs that are likely essential for viability and 4new HPKs that have important function in fruiting body for-mation or spore germination.

MATERIALS AND METHODS

Identification and phylogenetic analysis of TCS proteins. TCS proteins cangenerally be divided into three groups: HPKs, which contain either a HisKA,HisKA_2, HisKA_3 (hereafter these three domains will be collectively referredto as the HisKA domain), or Hpt domain and the conserved HATPase_c do-main; HPK-like proteins, which contain the HisKA domain and lack theHATPase_c domain or vice versa; and RRs, which contain the conserved re-ceiver domain. HPK, HPK-like proteins, and receiver domain-containing pro-teins were identified using the TIGR Comprehensive Microbial Resource data-base (http://cmr.tigr.org/tigr-scripts/CMR/CmrHomePage.cgi) and HMMER(HMMER 2.3; http://selab.janelia.org/) searches on the M. xanthus proteomeusing the pfam HisKA, HisKA_2, HisKA_3, Hpt, and HATPase_c matrices fordomains in HPKs and the Response_reg matrix for receiver domains. The E-value cutoffs were adopted from the SMART web service domain identificationtool (51) and were 0.02 for the HisKA, Hpt, and Response_reg domains and 1e-5for the HATPase_c domain. Sequences from proteins containing one or more ofthe homologous matrices of HisKA, Hpt, HATPase_c, and Response_reg wereextracted based on boundaries identified using the HMMER results, SMARTanalysis results (36), and manual curation using the domain boundary definitionsdescribed by Grebe and Stock (15). To generate phylogenetic trees, the M.xanthus TCS domains were compared to the domain sets used by Grebe andStock (15) by aligning all HPK domains and, separately, all receiver domainsusing the CLUSTALW 1.83 alignment software (gap open penalty � 50; gapextension penalty � 0.5; Gonnet protein matrix) (http://www.ebi.ac.uk/clustalw)(62). The resulting alignments were improved by manual curation and were thenused to generate phylogenetic trees using the neighbor-joining method (49)(1,000 bootstrapping replications implemented).

To identify output domains in RR, HMMER searches were performed on allM. xanthus proteins bearing receiver domains using all matrices provided in thepfam database (http://pfam.sanger.ac.uk/) (10). Results from the searches wereinspected, and based on the annotations provided for each pfam matrix, all hitsfrom known output domains were noted. For those proteins for which the abovesearch failed to identify an output domain, possible cryptic, conserved domainswere searched for as follows: peptide sequences leading and trailing the receiverdomains from these proteins were collected and subjected to a BLAST search ofthe NCBI nonredundant protein database. Results were manually analyzed, butno new domains were identified.

Transmembrane helices were predicted using the TMHMM 2.0c softwarepackage with the provided default model and options (http://www.cbs.dtu.dk/services/TMHMM) (56). If a protein was predicted to contain only a single

614 SHI ET AL. J. BACTERIOL.

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

transmembrane helix, the location of the transmembrane helix was inspected. Ifthe helix was located within the first 20 amino acid residues and would thusoverlap with the signal peptide, the HPK was categorized as cytoplasmic.

Classification of TCS proteins in M. xanthus based on genetic organization.Based on the genetic organization of TCS genes, a set of criteria was developedto classify these genes into three groups (Fig. 1), as follows.

Complex gene clusters. Complex gene clusters were defined as gene clusterscontaining two or more RR genes, independent of their transcriptional direction,gene clusters containing two or more HPK or HPK-like genes, independent oftheir transcriptional direction, and gene clusters containing three or more TCSgenes, independent of their transcriptional direction.

Paired genes. Paired genes were defined as two adjacent genes encoding anHPK or an HPK-like protein and an RR and transcribed in the same direction.

Orphan genes. All other gene organizations were considered orphan genes.Three che gene clusters are flanked by three orphan genes (MXAN6953,MXAN6955, and MXAN6966) encoding HPKs and one gene (MXAN2687)encoding an HPK-like protein that are not CheA-like, and these genes areclassified as orphan.

Cell growth and development for DNA microarray analysis. M. xanthusDK1622 wild-type cells were grown in liquid 1% CTT medium (18) at 32°C to adensity of 5 � 108 cells/ml, harvested, and resuspended in prewarmed (32°C)MC7 (10 mM morpholinepropanesulfonic acid [pH 7.0], 1 mM CaCl2) to acalculated density of 5 � 109 cells/ml. The cell suspension was diluted 1:8 withMC7 and 35 ml of the resulting suspension transferred to a sterile 145-mmculture dish (Greiner Bio-One) and incubated at 32°C. After 0, 2, 4, 6, 9, 12, 15,18, or 24 h of development, cells were harvested, and immediately frozen inliquid nitrogen and stored at �80°C. For the preparation of total RNA fromexponentially growing cells, cells were grown as described above, harvested at adensity of 5 � 108 cells/ml, and immediately frozen in liquid nitrogen and storedat �80°C. Total RNA from these cells served as “reference RNA.” Threebiological replicate time course experiments were performed.

Isolation of total RNA and DNase I treatment. Total RNA was isolated fromcell pellets using the hot-phenol method (44). One hundred micrograms totalRNA was treated with 20 U RNase-free DNase I (Fermentas) for 60 min at 37°C.RNA was purified using the RNeasy minikit (Qiagen). The absence of DNA wasverified by PCR. RNA quality was checked using agarose gel electrophoresis(50).

cDNA synthesis, fluorescent labeling, and hybridization. cDNA synthesis wascarried out using 25 �g of DNA-free total RNA for each experimental sample(RNA from developing cells) and 25 �g of the DNA-free reference sample asdescribed previously (20, 44) with the following modifications. cDNA synthesis

was carried out in the presence of 0.5 mM (each) dATP, dCTP, and dGTP, 0.1mM dTTP, and 0.4 mM aminoallyl-dUTP in a total volume of 30 �l. Subse-quently, RNA was hydrolyzed by addition of 10 �l of 1 M NaOH and 10 �l of 0.5M EDTA and incubated at 65°C for 15 min, followed by addition of 10 �l of 1M HCl. cDNA was purified using a Zymo kit (Zymo Research), vacuum dried,and resuspended in 13 �l of fresh 100 mM sodium bicarbonate, pH 9. cDNA ofthe reference probe was labeled with Cy5 and cDNA of the developmental probewith Cy3 as described previously (20). The M. xanthus DNA microarrays cover88% of all protein-coding genes in the M. xanthus genome and were printed atthe Stanford DNA microarray printing facility and postprocessed as describedpreviously (8, 44). Briefly, each open reading frame (ORF) is represented on themicroarray as a 275- to 325-bp PCR fragment. Hybridizations were carried out asdescribed previously (44).

Data acquisition and analysis. Microarrays were scanned simultaneously attwo wavelengths (for Cy3, 532 nm; for Cy5, 635 nm) using a GenePix 4000Bmicroarray scanner (Axon Instruments, Inc.). Image analysis and processing wasperformed using the GenePix Pro 6.0 software package (Axon Instruments, Inc.).The ratio-normalized data set (mean ratio of medians � 1) containing mediansignal intensity and median signal background from each channel was furtheranalyzed using Aquity 4.0 software (Axon Instruments, Inc.) and the SignificanceAnalysis of Microarrays (SAM) software (version 2.0), which assigns a score toeach feature on a microarray on the basis of changes in gene expression relativeto the standard deviation of repeated measurements (63). A filtered subset of allfeatures printed on the array was selected based on the following criteria: (i)found by the Genepix Pro 6.0 spot-finding algorithm (“Flags” � 0) and (ii)signal-to-noise ratio of either the Cy3 (532 nm) or Cy5 (635 nm) channel wasgreater than 2. For statistical significance analysis of the selected data points,SAM was used to calculate a t-like statistic (d) based on the estimated varianceof the data. The cutoff value of the SAM analysis was chosen as the value wherethe median false discovery rate became 0%. For the selected features, ratios wereaveraged and subjected to a 1.5-fold cutoff criterion (at least in one time pointduring development) and were further selected for the ones where data pointswere present for all developmental time points and the expression value at 0 hwas between �1.5 and �1.5 (not regulated).

Quantitative real-time PCR. Verification of gene expression data obtained byDNA microarray analysis was carried out using quantitative real-time PCR(qRT-PCR) as described for M. xanthus (44). Briefly, cDNA was synthesized withthe cDNA Archive kit (ABI) from 1.0 �g of DNA-free total RNA from sixdifferent developmental time points (0, 6, 12, 18, and 24 h) from one biologicalexperiment from the DNA microarray analyses. Primers for qRT-PCR (seeTable S1 in the supplemental material) were designed with Primer Express 2.0(ABI) to give fragments with sizes of 60 to 150 bp. The qRT-PCRs were carriedout in triplicate in a total volume of 25 �l containing 12.5 �l SYBR green PCRmaster mix (ABI), 1 �l of each primer (10 �M), 0.1 �l cDNA, and 11.9 �l H2O.qRT-PCRs were performed on an AB 7300 real-time PCR detection systemusing standard conditions. Expression ratios were calculated as the absoluteexpression level in developing cells over the absolute expression level in vegeta-tive cells. The efficiency of each primer pair was determined using four differentconcentrations of DK1622 chromosomal DNA (10 ng/�l, 1.0 ng/�l, 0.1 ng/�l, and0.01 ng/�l) as a template in qRT-PCRs.

Construction of mutants in genes coding for HPKs. In-frame deletion mutantswere constructed by two-step homologous recombination (Fig. 2). Briefly, ap-proximately 1,100-bp PCR products containing the in-frame deletions werecloned in the plasmid pBJ114 (22), which contains the galK gene for counters-election. Primers used for the constructions are listed in Table S2 in the supple-mental material. Four primers (A, B, C, and D) were designed to amplify the1,100-bp fragment carrying an hpk in-frame deletion by PCR with M. xanthuschromosomal DNA as a template. Briefly, primers A and B were used to amplifythe upstream flanking region of the hpk gene. Primer A contained a restrictionsite for cloning in pBJ114, and primer B contained either a restriction site forcloning or a region complementary to the downstream flanking PCR fragment.Primers C and D were used to amplify the downstream flanking fragment of thehpk gene. Primer C contained either a restriction site for cloning or a regioncomplementary to the upstream flanking PCR fragment, and primer D containeda restriction site for cloning in pBJ114. The fragments AB and CD were used togenerate the full-length in-frame deletion fragment either by direct cloning or ina second PCR with primers A and D and the two flanking PCR fragments astemplates. The hpk in-frame fragments were cloned in the plasmid pBJ114 andtransformed into E. coli Top 10 (Invitrogen) (F� mcrA �(mrr-hsdRMS-mcrBC)�80lacZ�M15 �lacX74 recA1 ara�139 �(ara-leu)7697 galU galK rpsL endA1nupG) and checked by sequencing. Correct plasmids were introduced into the M.xanthus wild-type strain DK1622 by electroporation (25). The insertion of plas-mids after the first homologous recombination was confirmed by PCRs with

FIG. 1. Classification scheme for two-component system genes.Schematic diagram of classification schemes for two-component sys-tem genes. The definition of paired and orphan TCS genes includesinformation about transcription direction as indicated by the arrowsymbols. Complex TCS gene clusters include clusters containing two ormore RR genes, clusters containing two or more HPK or HPK-likegenes, and clusters with three or more TCS genes irrespective oftranscription direction, as indicated by the box symbols. For complexgene clusters, only the most common gene organizations are shown(see Tables S3, S4, and S5 in the supplemental material for all geneorganizations found in these clusters).

VOL. 190, 2008 TWO-COMPONENT SYSTEMS IN M. XANTHUS 615

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

three primer pair combinations (primers are listed in Table S2 in the supple-mental material): primers E (binds upstream of primer A) and F (binds down-stream of primer D), primers E and M13-forward (hybridizes to pBJ114), andprimers F and M13-reverse (hybridizes to pBJ114). For each in-frame construct,at least one clone with the insertion of the plasmid in the upstream flankingregion of the hpk gene and one clone with the insertion in the downstreamflanking region of the hpk gene were chosen for the second homologous recom-bination. To isolate clones containing the in-frame deletion, cells were plated onCTT plates with 1% or 2% galactose (Sigma) for counterselection. Galactose-resistant and kanamycin-sensitive colonies were screened out and checked by twoPCRs with the primers E and F and the primers G and H, which bind to the

deleted part of the hpk gene, to verify the in-frame deletion. For MXAN3036 andMXAN4988, no in-frame deletions were obtained with the 1,100-bp in-framedeletion constructs using the primers labeled “short” in Table S2 in the supple-mental material, i.e., all galactose-resistant clones contained in the intact genes.Therefore, for both genes in-frame deletion constructs on 1,400-bp fragmentswere generated using the primers labeled “long” in Table S2 in the supplementalmaterial. Also, with these constructs, no in-frame deletions were isolated. In thecase of MXAN0060, primers A and B were used to generate an internal fragmentof MXAN0060, which was cloned into pBGS18 (57) to generate pXS011, whichwas used to generate an insertion mutation in MXAN0060. M. xanthus strains arelisted in Table 1.

Development of M. xanthus, spore assays, and motility tests. To test fordevelopmental defects, M. xanthus cells were grown as described above anddeveloped as described previously in submerged culture, on clone fruiting (CF)agar or on TPM agar (55). Spore numbers were determined as the number ofspores formed after 72 h and 120 h of starvation by harvesting 5 � 108 cells fromCF agar. Cells were placed for 2 h at 50°C and briefly sonicated to dispersefruiting bodies. Spores were counted in a hemacytometer (depth, 0.1 mm;Marienfeld). Three biological experiments were performed for each strain todetermine the sporulation efficiency. To determine the number of germinatingspores, spore solutions were plated on 1.0% CTT agar plates. The number offruiting bodies formed was determined after 120 h of starvation by manuallycounting fruiting bodies formed in three 20-�l cell aliquots on CF agar in threebiological experiments. The surface area covered by individual fruiting bodieswas calculated by measuring the total area of fruiting bodies formed in three20-�l cell aliquots on CF agar after 120 h of starvation in three biologicalexperiments using the area measurement tool in Metamorph (Molecular De-vices), followed by division by the total number of fruiting bodies formed.

To test for motility defects, cells were grown as described above and plated on0.5% CTT medium supplemented with 0.5% or 1.5% agar as described previ-ously (52).

Microarray data accession number. The microarray data discussed in thispublication have been deposited in the NCBIs Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) and are accessible through Gene ExpressionOmnibus Series accession number GSE9477.

RESULTS

Identification of TCS genes in M. xanthus. We analyzed theM. xanthus genome and identified 272 genes that encode pro-

TABLE 1. M. xanthus strains used in this work

Strain Genotype Reference

DK1622 Wild type 24MS1512 �MXAN1014 (� �sdek) 46PH1017 �MXAN6855 (� �espC) B. Lee and P. Higgsa

SA2107 MXAN0060::pXS011 This studySA2112 �MXAN3290 This studySA2115 �MXAN6315 This studySA2117 �MXAN0340 This studySA2118 �MXAN0736 This studySA2119 �MXAN7123 This studySA2120 �MXAN0928 This studySA2121 �MXAN6994 This studySA2124 �MXAN2763 This studySA2126 �MXAN5483 This studySA2130 �MXAN3098 This studySA2131 �MXAN0571 This studySA2133 �MXAN7398 This studySA2134 �MXAN4465 This studySA2136 �MXAN0176 This studySA2137 �MXAN6941 This studySA2138 �MXAN0712 This studySA2139 �MXAN0931 (� �espA) This studySA2142 �MXAN0706 This studySA2143 �MXAN7206 (� �mokA) This studySA2144 �MXAN6996 (� �asgD) This study

a Max Planck Institute for Terrestrial Microbiology.

FIG. 2. Outline of strategy for generating in-frame deletions ingenes encoding histidine protein kinases. For each gene to be deleted,a plasmid was constructed by PCR and cloning with approximately550-bp regions of homology upstream (light gray) or downstream(dark gray) flanking the in-frame construct. In a two-step procedure,deletion strains were isolated by first selecting for kanamycin resis-tance, followed by galactose counterselection using the galK gene car-ried on the plasmid and screening for loss of kanamycin resistance.Cells harboring galK die in the presence of galactose. Candidate clonesfor carrying the in-frame deletion were identified as Galr and KanS.Clones carrying the in-frame deletion were identified by PCR using theprimer pairs EF and GH. For simplicity, the first homologous recom-bination is shown only for the recombination occurring in the upstreamflanking region.

616 SHI ET AL. J. BACTERIOL.

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

teins of TCS (Materials and Methods). Among these genes, 21are localized in 8 loci that encode proteins which are part ofChe-like systems, i.e., these clusters encode homologues ofChe proteins and the HPK has a domain structure similar tothat found in CheA (See Table S3 and S4 in the supplementalmaterial for a list of these genes). In this report, we specificallyfocus on the 251 TCS genes encoding proteins that are not partof Che-like systems. These 251 genes include 118 HPK, 119RR, and 14 HPK-like genes (Table 2). HPK genes are pre-dicted to encode bona fide kinases which contain the conservedHATPase_c domain and a HisKA domain with the phosphor-ylatable His residue. RR genes are predicted to encode bonafide RRs which either consist of a single receiver domain or aremultidomain proteins which in addition to the receiver domainalso contain an output domain. HPKs that contained one ormore receiver domains were classified as hybrid HPK. HPK-like genes encode proteins that either contain a HisKA domainwith the phosphorylatable His residue and lack the HATPase_cdomain or vice versa. The HATPase_c domain is also found inother ATPases, such as DNA gyrase B and the DNA repairprotein MutL (11). To avoid annotating such proteins as HPK-like proteins, proteins were classified as HPK-like proteinsonly if they were encoded by genes located next to a TCS gene(e.g., MXAN0461 and MXAN4203) or if they contained one ormore receiver domains (e.g., MXAN0230 and MXAN4432)(See Table S5 in the supplemental material for the domainstructures of these proteins).

Given the large number of TCS genes in M. xanthus, we nextasked whether the kinase and receiver domains belong to theestablished kinase and receiver families as defined by Grebeand Stock (15). To analyze the relationship between the M.xanthus TCS proteins, we aligned all kinase domains (encom-passing the HisKA and HATPase_c domains) and receiver

domains in RR, including receiver domains in hybrid HPKsand in HPK-like proteins, with those used by Grebe and Stock(15), and phylogenetic trees were generated. Analysis of thealignments and trees showed that all the kinase and receiverdomains in the M. xanthus TCS proteins belong to the estab-lished Grebe and Stock families (See Table S3 and Table S4 inthe supplemental material for family assignment). Thus, de-spite the large number of TCS proteins, there is no evidencethat M. xanthus has evolved new kinase or receiver domains.Rather, M. xanthus appears to use domains also found in otherspecies.

Genetic organization of two-component genes in M. xanthus.During the analysis of M. xanthus TCS genes, we noticed thatmany of these genes are not organized as pairs as is typicallyreported. To analyze the genetic organization of TCS genes,we developed a set of criteria according to which TCS geneswere divided into three categories: complex gene clusters, or-phan genes, and paired genes (Fig. 1) (see Materials and Meth-ods). The genetic organization of TCS genes in M. xanthusdiverges significantly from the standard paired organization(Table 2; Fig. 3, 4, and 5): 55% (138 genes out of 251 total) ofTCS genes are orphans, 16% (39 out of 251) of TCS genes arelocated in complex gene clusters, and only 29% (74 out of 251)are found as paired genes.

Histidine protein kinases. As a first step in understandinghow M. xanthus TCS proteins are connected, we divided the118 HPKs into HPKs and hybrid HPKs, which in addition tothe HisKA and HATPase_c domains also contain one or morereceiver domains. We identified 31 hybrid kinases (26% of thetotal number of HPKs), which contain between 1 and 3 re-ceiver domains (Table 2; Fig. 3) (see Table S3 in the supple-mental material for detailed domain organization). The distri-bution of the 31 hybrid HPKs is highly biased. Eight of the

TABLE 2. Summary of two-component system genes and proteins in M. xanthus

Category Total no. of genes No. of complex geneclusters No. of orphan genes No. of paired genes

HPK genes/proteins 118 (31 hybrid) 18 (8 hybrid) 68 (22 hybrid) 32 (1 hybrid)RR genes/proteins 119 18 64 37HPK-like genes/proteins 14 3 6 5Total TCS genes/proteins 251 39 138 74

FIG. 3. Organization diagram of histidine protein kinase genes and proteins in the M. xanthus genome. HPK genes were divided into threecategories based on genetic organization. The number of hybrid HPKs is indicated in parentheses for each category. In the second layer, HPKsare divided according to their likely subcellular localization based on the presence (integral membrane) or absence (cytoplasmic) of transmem-brane helices. In the third layer, HPK genes are divided according to transcriptional regulation during development. Note that in the third layer,not all HPK genes are included because they are not all present on the M. xanthus DNA microarray or they were not tested by qRT-PCR.

VOL. 190, 2008 TWO-COMPONENT SYSTEMS IN M. XANTHUS 617

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

hybrid HPKs are encoded by complex gene clusters (corre-sponding to 44% of all HPKs encoded by these genes), 22 ofthe hybrid HPKs are encoded by orphan genes (correspondingto 32% of all HPKs encoded by these genes), and only 1 hybridHPK is encoded by a paired gene (corresponding to 3% of allHPKs encoded by these genes). Among the 31 hybrid HPKs,only 1, MXAN2317, is predicted to contain an Hpt domain.

HPKs are generally described as being integral membraneproteins (58). To determine whether HPKs in M. xanthus con-form to this general description, we analyzed HPKs for thepresence or absence of transmembrane helices. Among the 118HPKs, 45 (38% of the total number of HPKs) are likely to beintegral membrane proteins and 73 are likely to be cytoplasmic(Fig. 3) (see Table S3 in the supplemental material for detailedresults). Also in this analysis, we found a biased distribution ofthe two types of HPKs. Thus, 11 of the predicted cytoplasmicHPKs are encoded by complex gene clusters (corresponding to61% of all HPKs encoded by these genes), 54 of the predictedcytoplasmic HPKs are encoded by orphan genes (correspond-ing to 79% of all HPKs encoded by these genes), and only 8 of

the predicted cytoplasmic HPKs are encoded by paired genes(corresponding to 25% of all HPKs encoded by these genes).

RRs and output domains. As a second step in understandingthe connectivity of TCS in M. xanthus, we divided the RRs intosingle-domain RRs, which consist only of the receiver domain,and multidomain RRs, which in addition to the receiver do-main contain an output domain. The 119 RRs can be dividedinto 38 without (32% of the total number of RRs) and 81 withoutput domains (Table 3; Fig. 4) (see Table S4 in the supple-mental material for the detailed domain organization). All chegene clusters contain either a CheAY hybrid kinase or CheYhomologs (see Tables S3 and S4 in the supplemental materialfor the detailed description of proteins encoded by che geneclusters). Thus, the remaining single-domain RRs are notlikely to be CheY paralogs.

The distribution of RRs without output domains is highlybiased. The complex gene clusters include 18 RR genes en-coding 12 RR with and 6 RR without output domains (corre-sponding to 33%). The output domains comprise five DNAbinding domains, five GGDEF domains, which are involved in

FIG. 4. Organization diagram of response regulator genes and proteins in the M. xanthus genome. RR genes were divided into three categoriesbased on genetic organization. In the second layer, RRs are divided according to the presence or absence of output domains. In the third layer,RRs are divided according to transcriptional regulation during development. Note that in the third layer, not all RR genes are included becausethey are not all present on the M. xanthus DNA microarray or they were not tested by qRT-PCR.

FIG. 5. Organization diagram of histidine protein kinase-like genes and proteins in the M. xanthus genome. Genes encoding HPK-like proteinswere divided into three categories based on genetic organization. In the second layer, HPK-like proteins are divided according to their likelysubcellular localization based on the presence (integral membrane) or absence (cytoplasmic) of transmembrane helices. Note that in the thirdlayer, not all HPK-like genes are included because they are not all present on the M. xanthus DNA microarray or they were not tested by qRT-PCR.

618 SHI ET AL. J. BACTERIOL.

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

cyclic-di-GMP synthesis and regulation (21, 47), and two do-mains of unknown function (DUF). The 64 orphan RR genesinclude 35 encoding RR with and 29 without output domains(corresponding to 45%). The four largest categories of outputdomains for the orphan RR comprise 13 DNA binding do-mains, five GGDEF domains, two PilZ domains, which bindcyclic-di-GMP (2, 48), and 10 DUF. Finally, for the 37 RRencoded by paired genes, 34 contain an output domain andonly 3 are without an output domain (corresponding to 9%).The 34 output domains comprise 32 DNA binding domainsand 2 DUF. The distribution of the output domains suggeststhat the two main outputs from TCS in M. xanthus are generegulation (50 domains) and cyclic-di-GMP regulation (12 do-mains). RRs with DUF are overrepresented among RRs en-coded by complex gene clusters and orphan genes.

Transcriptional regulation of TCS genes during develop-ment. Thirty-five TCS genes have been identified as importantfor fruiting body formation and sporulation (see Table S3,Table S4, and Table S5 in the supplemental material for thespecific genes). Several of these genes are developmentallyregulated at the transcriptional level during development.Thus, we reasoned that one approach to identify TCS proteinsthat may have a function in fruiting body formation would beto analyze the expression profiles of TCS genes during devel-opment.

For these experiments, we used an M. xanthus DNA mi-croarray covering 88% of the 7,380 ORFs on the M. xanthusgenome (see Materials and Methods). The detailed analysis ofthe experiments will be described elsewhere (N. Hamann, S.Wegener-Feldbrugge, L. Søgaard-Andersen, and R. Hed-derich, data not shown). Total RNA was isolated from mid-exponentially growing wild-type cells (DK1622) and from dif-ferent time points during development (0, 2, 4, 6, 9, 12, 15, 18,and 24 h), and cDNA was prepared, labeled with Cy3 (samplesfrom development) and Cy5 (reference), and competitivelyhybridized to the microarray. As a reference, the RNA isolatedfrom mid-exponentially growing wild-type cells was used. Atotal of three biological experiments were performed. Thus,data analysis (Materials and Methods) was carried out on threeexperimental values for each gene. The ratio-normalized dataset was analyzed using Acuity 4.0 software (Axon Instruments)and SAM software, version 2.0, which assigns a score to eachfeature on a microarray on the basis of changes in gene ex-pression relative to the standard deviation of repeated mea-surements (63). Genes called to be significantly regulated dur-ing development were selected by a delta value of the SAM

analysis where the false discovery rate became 0% in combi-nation with a 1.5-fold cutoff and data points for all time points.

Among the 200 TCS genes present on the array, 50 displayedaltered expression during development (Fig. 3, 4, and 5; seeTable S3, Table S4, and Table S5 in the supplemental materialfor details on the expression of individual genes). Develop-mentally regulated genes exhibited expression ratios in therange of 10.9-fold up-regulation to 6.6-fold down-regulation.The changes in gene expression during development wereasymmetric: 46 genes were up-regulated, and 4 genes weredown-regulated. To validate the significance of the expressiondata obtained from the DNA microarrays, qRT-PCR was ap-plied to 11 genes (10 genes up-regulated during developmentand 1 gene not showing regulation). The transcriptional dif-ferences determined in the microarray experiments were con-firmed by the qRT-PCR analysis (see Table S3, Table S4, andTable S5 in the supplemental material for details of qRT-PCRdata).

To generate a complete data set on the expression profile foran entire category of TCS genes, we tested by qRT-PCR theexpression during development of the 13 orphan HPK genesfor which no expression data were available from the DNAmicroarray experiments. Among these genes, we found that sixwere up-regulated and seven were down-regulated during de-velopment (see Table S3 in the supplemental material fordetails of the qRT-PCR data).

In total, we identified 63 TCS genes out of 213 TCS genestested either in microarray analyses or by qRT-PCRC thatwere transcriptionally regulated during development (Fig. 3,Fig. 4 and Fig. 5) (see Tables S3, S4, and S5 in the supplemen-tal material for the expression of individual genes). Fifty-twogenes were up-regulated and 11 genes down-regulated. Thirty-six percent and 13% of the tested orphan and complex genes,respectively, were transcriptionally regulated during develop-ment, whereas only 12% of the tested paired genes were tran-scriptionally regulated during development.

Genetic analysis of transcriptionally up-regulated, orphanhistidine protein kinase genes. The gene expression profilingexperiments indicate that orphan TCS may have importantfunctions in fruiting body formation. To test this hypothesisand to potentially identify novel TCS genes required for de-velopment, we focused on the 25 orphan HPK genes that aretranscriptionally up-regulated during development (Fig. 3)(Table 4 contains a list of all 25 genes, including expressiondata). These 25 genes include espA (7), sdeK (13), espC (34),asgD (6), and mokA (28), which have previously been sug-gested to be important for development. To analyze the im-portance of the remaining 20 genes for fruiting body forma-tion, we sought to generate in-frame deletions in 19 of thesegenes using a two-step recombination procedure (Fig. 2); forMXAN0060, an insertion mutant was constructed. In-framedeletions were preferred over insertion mutants to avoid polareffects on downstream genes. In addition, we generated orobtained in-frame deletions of the five previously identifiedorphan HPKs important for development in order to system-atically compare developmental defects. For each of the genesMXAN3036 and MXAN4988, more than 200 galactose-resis-tant, kanamycin-sensitive clones were tested (see Materialsand Methods). For both genes, all colonies tested containedthe intact HPK gene. These observations suggest that

TABLE 3. Summary of output domains in RRs encoded in theM. xanthus genome

Output domaincategory

Total no. ofRRs withdomains

No. of RRs encoded by:

Complex geneclusters

Orphangenes

Pairedgenes

None 38 6 29 3DNA binding 50 5 13 32GGDEF 10 5 5 0PilZ 2 0 2 0Other domainsa 5 0 5 0DUF 14 2 10 2

a This group consists of various output domains present only once.

VOL. 190, 2008 TWO-COMPONENT SYSTEMS IN M. XANTHUS 619

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

MXAN3036 and MXAN4988 are essential genes for viability.For the remaining 22 genes, we obtained stable in-frame de-letion mutants. For MXAN0060, we obtained a stable insertionmutant.

We next examined the phenotypes of the wild-type strain(DK1622) and the 23 HPK mutants. To test for motility de-fects, cell spreading was examined on 0.5% CTT medium con-taining 0.5% agar, which favors motility by means of the S-motility system, or 1.5% agar, which favors motility by meansof the A-motility system (52). None of the 23 mutants dis-played motility defects on these two types of surfaces (data notshown). To test the mutants for developmental defects, cellswere exposed to starvation under three different conditions,i.e., CF starvation agar, TPM starvation agar, and in sub-merged culture and the sporulation frequencies determinedafter 72 h and 120 h of starvation on CF agar. Moreover, levelsof germinating spores were determined after 72 h and 120 h ofstarvation on CF agar. Individual mutants displayed similarphenotypes under all three developmental conditions (datafrom development on CF agar and in submerged culture areshown in Fig. 6 and Table 5). DK1622 wild-type cells hadformed fruiting bodies at 24 h, and at 48 h the fruiting bodieshad darkened (Fig. 6; Table 5). Sixteen mutants displayednormal development, sporulation, and spore germination (datanot shown). As previously reported (13, 46), MS1512 carrying�MXAN1014 (� �sdek) was unable to aggregate to formfruiting bodies and was strongly reduced in sporulation. Alsoas previously reported, a mutant carrying �MXAN0931 (�

�espA) (SA2139) displayed early aggregation with the forma-tion of many small and irregularly shaped fruiting bodies; also,SA2139 displayed early sporulation with many spores localizedoutside the fruiting bodies (Fig. 6B). SA2144 carrying�MXAN6996 (� �asgD) displayed delayed aggregation butnormal levels of sporulation. This is in disagreement with aprevious report, in which an insertion in asgD was reported tocause aggregation as well as sporulation defects (6). We at-tribute these differences to differences in strain backgroundsand mutations being analyzed in the previous report and thedata presented here. SA2112 carrying �MXAN3290 displayeddelayed aggregation and formation of abnormally shaped fruit-ing bodies. SA2112 sporulated at wild-type levels, however,many of the spores were localized outside fruiting bodies.SA2134 carrying �MXAN4465 displayed normal timing of ag-gregation and sporulation. However, this mutant formed 1.7-fold � 0.4-fold more fruiting bodies than the wild type andindividual fruiting bodies were smaller than those formed bythe wild type, covering an area of only 60% � 7% of that of awild-type fruiting body. Moreover, the fruiting bodies formedby SA2134 in submerged culture were less condensed thanthose formed by the wild type (Fig. 6B). SA2138 carrying�MXAN0712 was unable to aggregate to form fruiting bodiesand was strongly reduced in sporulation. Finally, SA2118 car-rying �MXAN0736 displayed normal aggregation and fruitingbody formation and sporulated at wild-type levels. However,these spores germinated at a level threefold lower than thatobserved with the wild type. It should be noted that the two

TABLE 4. Orphan histidine protein kinase genes in M. xanthus transcriptionally up-regulated during development

TIGR_MXAN(gene symbol)

Max. expression ratio (microarrays)/induction time (h)a,b

Max. expression ratio (qRT-PCR)/induction time (h)a,b

No. of TMhelicesc Domain organizationd

MXAN0060 NOA 19.3�-up/0–6 2 HisKA-HATPase_cMXAN0176 NOA 14.8�-up/6–12 0 HisKA-HATPase_cMXAN0340 2.7�-up/4–6 ND 0 HisKA-HATPase_cMXAN0571 1.9�-up/2–4 ND 0 HisKA-HATPase_cMXAN0706 NOA 3.4�-up/0–6 0 HisKA-HATPase_cMXAN0712 1.6�-up/12–15 ND 0 HisKA-HATPase_c-RR-RRMXAN0736 2.2�-up/0–2 128�-up/0–6 2 HisKA-HATPase_cMXAN0928 10.9�-up/2–4 29.7�-up/0–6 0 HisKA-HATPase_cMXAN0931 (� espA) 4.4�-up/0–2 ND 0 HisKA-HATPase_c-RRMXAN1014 (� sdeK) 2.6�-up/0–2 ND 0 HisKA-HATPase_cMXAN2763 2.4�-up/0–2 ND 0 HisKA-HATPase_c-RRMXAN3036 2.1�-up/4–6 29.8�-up/0–6 3 HisKA-HATPase_cMXAN3098 1.9�-up/0–2 4.1�-up/0–6 0 HisKA-HATPase_cMXAN3290 2.8�-up/0–2 4.4�-up/0–6 0 HisKA-HATPase_c-RRMXAN4465 2.1�-up/4–6 ND 0 HisKA-HATPase_c-RRMXAN4988 1.6�-up/0–2 2.8�-up/0–6 8 HisKA-HATPase_cMXAN5483 2.2�-up/4–6 ND 0 HisKA-HATPase_cMXAN6315 4.2�-up/0–2 36.5�-up/0–6 0 HisKA-HATPase_c-RRMXAN6855 (espC) 1.7�-up/12–15 ND 8 HisKA-HATPase_c-RRMXAN6941 NOA 7.5�-up/6–12 0 HisKA-HATPase_cMXAN6994 4.4�-up/2–4 76.4�-up/0–6 0 HisKA-HATPase_cMXAN6996 (asgD) 1.8�-up/12–15 ND 0 RR-HisKA-HATPase_cMXAN7123 4.0�-up/2–4 143�-up/0–6 0 HisKA-HATPase_cMXAN7206 (mokA) NOA 9.2�-up/0–6 0 HisKA-HATPase_c-RRMXAN7398 NOA 3.8�-up/0–6 0 HisKA-HATPase_c

a Expression ratios were calculated as the expression in developing cells over the expression in vegetative cells. Maximum (max.) expression ratios indicate themaximum expression ratio for a particular gene during all time points tested using DNA microarrays (0, 2, 4, 6, 9, 12, 15, 18, and 24 h) and qRT-PCR (0, 6, 12, 18,and 24 h). Induction time indicates the time interval in which the expression ratio began to change relative to that observed in vegetative cells. “-up” indicatesup-regulation.

b NOA, gene not represented or not detected on DNA microarray; ND, not determined.c TM, transmembrane.d All domains present in a protein are indicated except for domains in the sensor domain.

620 SHI ET AL. J. BACTERIOL.

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

mutants carrying in-frame deletions of espC (PH1017) andmokA (SA2143) developed in a manner indistinguishable fromthat of the DK1622 wild type under all conditions tested andsporulated at wild-type levels and with the formation of ger-mination-proficient spores (data not shown). Inactivation ofthese two genes has previously been reported to cause devel-opmental defects (28, 34). We attribute the difference betweenpublished results and our results to differences in strain back-grounds used and to different mutations being analyzed.

DISCUSSION

Here we report the identification of 251 TCS genes, whichare not part of che gene clusters, in the M. xanthus genome.These 251 TCS genes make up 3.4% of the 7,380 ORFs (14) inthe M. xanthus genome. To our knowledge, this number ofTCS genes is the highest reported so far for any organism.However, when corrected for genome size, the number of TCSgenes in M. xanthus falls within that reported for other bacteriaand thus complies with the general rule that the number ofTCS proteins per genome increases with the square of thegenome size (11, 12). The number of TCS genes in M. xanthusis also in line with the general notion that growth of bacterialgenome size is accompanied by accumulation of paralogousprotein families.

The M. xanthus TCS genes could be divided into threeclasses based on their genetic organization. Fifty-five percentand 16% of all the TCS genes are organized as orphan genesor in complex gene clusters, respectively, and only 29% arefound as paired genes. We compared this genetic organizationof TCS genes to that in other bacteria and found that of the 59TCS genes in E. coli, 20% are orphan, 8% are in complex geneclusters, and 72% are paired; of the 63 TCS genes in B. subtilis,14% are orphan, 0% in complex gene clusters, and 86%paired; of the 86 TCS genes in Caulobacter crescentus, 63% areorphan, 7% in complex gene clusters, and 30% paired; of the126 TCS genes in Pseudomonas aeruginosa, 35% are orphan,8% in complex gene clusters, and 57% paired; and of the 186TCS genes in Streptomyces coelicolor, 25% are orphan, 9% incomplex gene clusters, and 66% paired (S. Huntley and L.Søgaard-Andersen, unpublished data). This comparison shows

FIG. 6. Developmental phenotypes of mutants containing in-frame deletions of orphan histidine protein kinase genes. (A) Developmentalphenotypes on CF agar. The indicated strains were starved on CF agar for the indicated periods of time. All strains analyzed are derivatives ofDK1622; below strain numbers, the in-frame deletion present in a particular strain is indicated. Scale bar, 1.0 mm. (B) Developmental phenotypesin submerged culture. The same strains as in panel A were exposed to starvation in submerged culture for 120 h. Scale bar, 100 �m.

TABLE 5. Sporulation frequencies of M. xanthus wild-type andHPK mutants

Strain Genotype

Sporulation frequency(%) at:a

72 h 120 h

DK1622 Wild type 66 � 4 100 � 6b

MS1512 �MXAN1014 (� �sdek) 0.001 0.001SA2139 �MXAN0931 (� �espA) 94 � 9 151 � 15SA2144 �MXAN6996 (� �asgD) 56 � 8 96 � 2SA2112 �MXAN3290 47 � 9 80 � 16SA2118 �MXAN0736 49 � 9 71 � 8c

SA2134 �MXAN4465 56 � 17 78 � 19SA2138 �MXAN0712 0.001 0.001

a Sporulation frequencies are presented relative to the sporulation level in thewild-type strain DK1622 after 120 h of starvation. Values are means and standarddeviations from three experiments.

b The absolute sporulation level of DK1622 was 14.7% � 1.0% at 120 h.c Spores of SA2118 germinated at a threefold-lower frequency than wild-type

spores.

VOL. 190, 2008 TWO-COMPONENT SYSTEMS IN M. XANTHUS 621

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

that the genetic organization of TCS genes shows large inter-species variations. Moreover, it is evident that the three cate-gories of TCS genes are not peculiar to the M. xanthus genome.

The genome size of M. xanthus is 9.14 Mb, and it has beensuggested that lineage-specific gene family expansions (LSE)were major contributors to the genomic expansion (14). Ourpreliminary analyses suggest that a large fraction of TCS genesin M. xanthus may have arisen by LSE (Huntley and Søgaard-Andersen, unpublished). Alm et al. (1) reported that delta-proteobacteria have a propensity for LSE of TCS genes. Inagreement with this, we find that the complements of TCSgenes in the three Myxococcales species Sorangium cellulosum(genome sequence provided by Rolf Muller, University ofSaarland, Saarbrucken, Germany), Stigmatella aurantiaca (ge-nome at http://cmr.tigr.org/tigr-scripts/CMR/CmrHomePage.cgi),and Anaeromyxobacter dehalogenans (genome at http://genome.jgi-psf.org/finished_microbes/anade/anade.home.html) also containa large fraction of TCS genes that likely arose by LSE (Huntleyand Søgaard-Andersen, unpublished). Interestingly, however, thecomplements of TCS genes that have expanded by LSE in thefour Myxococcales species appear to be different, suggesting thatfor each species the particular genes amplified provide that spe-cies with some selective benefits. A detailed description of LSE ofTCS genes in Myxococcales will be presented elsewhere (Huntleyand Søgaard-Andersen, unpublished).

We identified three groups of structurally remarkable TCSproteins in M. xanthus. One group consists of 14 HPK-likeproteins, which contain only a HisKA domain or a HATPase_cdomain (see Table S5 in the supplemental material). Four linesof evidence suggest that these genes are not pseudogenes butcode for functional proteins. First, three of these genes(MXAN0461 [� redE] [17], MXAN2670 [� asgA] [45], andMXAN5123 [� mrpA] [59, 60]) are required for development.Second, at least nine of the genes were found to be expressedin global transcriptional profiling experiments. Third, two ofthe genes are transcriptionally up-regulated during develop-ment. Finally, the HPK-like protein YojN, which functions in aphosphorelay with RcsB and RcsC (61) in Escherichia coli,provides evidence that HPK-like proteins may be functional. Asecond group of proteins with interesting structural featuresconsists of proteins that have organizations of signal transduc-tion domains, which have not been reported previously andwhich raise interesting questions in terms of how they functionin phosphotransfer reactions. For instance, the orphan HPKsMXAN2606 and MXAN2317 are predicted to have the do-main structures HisKA-HATPase_c-RR-HisKA-HATPase_cand HisKA-HATPase_c-RR-RR-Hpt, respectively, and theRR MXAN7362, which is encoded by a complex gene cluster,is predicted to have the domain structure RR-Hpt-RR-RR-GGDEF. The third group of TCS proteins with interestingstructural features consists of 14 RR with output DUF. TheseRRs are overrepresented among RRs encoded by complexgene clusters and orphan genes, i.e., 10 and 2 of these domainsare found in RRs encoded by orphan and complex genes,respectively. Interestingly, four of these proteins are orphanRRs involved in regulating gliding motility (MXAN2991 [�aglZ] [40, 67], MXAN4149 [� frzS] [39, 66], MXAN4461 [�romR] [35], and MXAN6627 [� sgnC] [68]).

The analysis of M. xanthus TCS proteins revealed structuralfeatures that have functional implications. First, 73 out of the

118 HPKs are predicted to be cytoplasmic, suggesting that themany HPKs in M. xanthus may be involved not in monitoringexternal stimuli or intercellular signals but rather in monitor-ing cytoplasmic stimuli. Alternatively, they could indirectly beinvolved in monitoring external stimuli or intercellular signalsby interacting with membrane proteins. Second, the analysis ofoutput domains in RRs suggests that the output responsesfrom TCS systems in M. xanthus center on three types, regu-lation of gene expression, regulation of cyclic-di-GMP metab-olism, and unknown functions.

Interestingly, we found strongly biased distributions of dif-ferent types of TCS proteins encoded by paired genes andorphan genes and in complex gene clusters. These biased dis-tributions have several functional implications, as discussedbelow. For paired TCS genes, the main implication is that alarge fraction of the corresponding proteins are part of simple1:1 TCS with an integral membrane HPK and a cognate RRthat is involved in regulation of gene expression. The predictedmembrane localization of the paired HPKs suggests that theyare primarily involved in monitoring external stimuli. More-over, the underrepresentation of these genes among transcrip-tionally regulated genes during development indicates thatthey may be functionally active in vegetative cells. Clearly, thelatter implication does not preclude a function during fruitingbody formation of these proteins. Consistently, 15 paired TCSgenes have been identified which are important for develop-ment (see Tables S3, S4, and S5 in the supplemental materialfor the identities of these genes).

For TCS proteins encoded by orphan genes or genes incomplex gene clusters, the biased distribution of protein char-acteristics and expression profiles suggests that the corre-sponding HPKs are primarily involved in monitoring cytoplas-mic stimuli (due to the overrepresentation of HPKs predictedto be cytoplasmic) and that the main output responses fromthe corresponding pathways are regulation of gene expression,regulation of cyclic-di-GMP metabolism, and unknown func-tions (as indicated by the overrepresentation of RRs with DUFoutput domains). Moreover, the overrepresentation of thesegenes among those that are transcriptionally regulated duringdevelopment suggests that many of these genes encode TCSproteins with a function only during development. Consis-tently, 16 orphan TCS genes (including the 4 identified in thisreport) and 6 TCS genes encoded in complex gene clustershave been shown to be important for development or sporegermination (see Tables S3, S4, and S5 in the supplementalmaterial for the identities of these genes). It should be notedthat transcriptional regulation during development does notpreclude a function in vegetative cells.

A question that remains to be addressed focuses on theconnectivity of the TCS proteins in M. xanthus. As mentionedfor the paired TCS genes, the almost complete absence ofhybrid HPKs and single-domain RRs in the correspondingproteins suggests that the paired genes encode proteins thatmake up simple, linear 1:1 pathways. For TCS proteins encodedby orphan genes and in complex gene clusters, the connectivityhas been analyzed experimentally only for the RedCDEF pro-teins (� MXAN0459 to MXAN0462), and the data suggest thatthese four proteins may constitute a complex phosphorelay (17).Since the connectivity of TCS proteins cannot be predicted basedon sequence conservation alone (54), this question, therefore,

622 SHI ET AL. J. BACTERIOL.

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

remains open for most of the TCS proteins encoded by orphangenes and in complex gene clusters. The close to 1:1 numericalratio of HPKs and RRs encoded by these genes could lead to thenotion that they could be organized in 1:1 pathways. However,two observations argue against this notion. First, hybrid HPKsare overrepresented among these proteins. Second, many ofthe RRs encoded by these genes are single-domain RRs. Theoverrepresentation of hybrid HPKs and RRs without outputdomains among the proteins encoded by complex gene clustersand orphan genes strongly suggests that the signal transductionpathways encoded by these genes are structured as phosphore-lays and/or are branched. Phosphorelays would likely dependon the presence of Hpt domain-containing proteins. In addi-tion to CheA kinases, we identified only two proteins contain-ing Hpt domains, the hybrid orphan HPK MXAN2317 and theRR MXAN7362, which is encoded in a complex gene cluster.It has been argues that Hpt domains are difficult to identify dueto the low level of sequence conservation (4); thus, M. xanthusmay indeed encode more proteins containing Hpt domains.Clearly, experimental analyses are needed to address the ques-tion of the connectivity of the M. xanthus TCS proteins.

We directly tested genetically the hypothesis that orphandevelopmentally up-regulated genes could be important fordevelopment by focusing on the 25 orphan HPK genes that areup-regulated at the transcriptional level during development.Among these genes, we found two (MXAN3036 andMXAN4988) that are likely to be essential for viability andseven that are important for development without having veg-etative defects. These seven genes include MXAN0931 (�espA) (7), MXAN1014 (� sdeK) (13, 46), and MXAN6996 (�asgD) (6), which have previously been shown to be importantfor development. In addition, we identified MXAN0712,MXAN0736, MXAN3290, and MXAN4465 as important fordevelopment or spore germination. Finally, inactivation ofMXAN6855 (� espC) (34) and MXAN7206 (� mokA) (28),which have previously been reported to be important for de-velopment, did not display developmental defects under ourconditions. How these seven proteins function in developmentremains to be determined. Clearly, the lack of developmentaldefects in the remaining 16 mutants could be caused by func-tional redundancy among HPKs. Nevertheless, our data havetwo implications: first, the transcriptional up-regulation of aTCS gene does not necessarily mean that this gene has anessential function during development (at least not under thethree conditions tested here). This notion is supported by theobservation that several other genes that are transcriptionallyup-regulated during development also do not have essentialfunctions during development (29). Second, even though TCSproteins clearly have important functions in development, thelarge number of TCS genes in M. xanthus may not have evolvedsolely to regulate fruiting body formation.

ACKNOWLEDGMENTS

We thank B. Lee, P. Higgs, and M. Singer for strains and R. Muller,University of Saarland, for access to the Sorangium cellulosum genomesequence prior to publication.

The International Max Planck Research School for Environmental,Cellular and Molecular Microbiology and the Max Planck Societysupported this work.

REFERENCES

1. Alm, E., K. Huang, and A. Arkin. 2006. The evolution of two-componentsystems in bacteria reveals different strategies for niche adaptation. PLoSComput. Biol. 2:e143.

2. Amikam, D., and M. Y. Galperin. 2006. PilZ domain is part of the bacterialc-di-GMP binding protein. Bioinformatics 1:3–6.

3. Appleby, J. L., J. S. Parkinson, and R. B. Bourret. 1996. Signal transductionvia the multi-step phosphorelay: not necessarily a road less traveled. Cell86:845–848.

4. Biondi, E. G., S. J. Reisinger, J. M. Skerker, M. Arif, B. S. Perchuk, K. R.Ryan, and M. T. Laub. 2007. Regulation of the bacterial cell cycle by anintegrated genetic circuit. Nature 444:899–904.

5. Burbulys, D., K. A. Trach, and J. A. Hoch. 1991. Initiation of sporulation inB. subtilis is controlled by a multicomponent phosphorelay. Cell 64:545–552.

6. Cho, K., and D. R. Zusman. 1999. AsgD, a new two-component regulatorrequired for A-signalling and nutrient sensing during early development ofMyxococcus xanthus. Mol. Microbiol. 34:268–281.

7. Cho, K., and D. R. Zusman. 1999. Sporulation timing in Myxococcus xanthusis controlled by the espAB locus. Mol. Microbiol. 34:714–725.

8. Diodati, M. E., F. Ossa, N. B. Caberoy, I. R. Jose, W. Hiraiwa, M. M. Igo, M.Singer, and A. G. Garza. 2006. Nla18, a key regulatory protein required fornormal growth and development of Myxococcus xanthus. J. Bacteriol. 188:1733–1743.

9. Fabret, C., V. A. Feher, and J. A. Hoch. 1999. Two-component signal trans-duction in Bacillus subtilis: how one organism sees its world. J. Bacteriol.181:1975–1983.

10. Finn, R., J. Mistry, B. Schuster-Bockler, S. Griffiths-Jones, V. Hollich, T.Lassmann, S. Moxon, M. Marshall, A. Khanna, R. Durbin, S. R. Eddy,E. L. L. Sonnhammer, and A. Bateman. 2006. Pfam: clans, web tools andservices. Nucleic Acids Res. 34:D247–D251.

11. Galperin, M. 2005. A census of membrane-bound and intracellular signaltransduction proteins in bacteria: bacterial IQ, extroverts and introverts.BMC Microbiol. 5:35.

12. Galperin, M. Y. 2006. Structural classification of bacterial response regula-tors: diversity of output domains and domain combinations. J. Bacteriol.188:4169–4182.

13. Garza, A. G., J. S. Pollack, B. Z. Harris, A. Lee, I. M. Keseler, E. F. Licking,and M. Singer. 1998. SdeK is required for early fruiting body developmentin Myxococcus xanthus. J. Bacteriol. 180:4628–4637.

14. Goldman, B. S., W. C. Nierman, D. Kaiser, S. C. Slater, A. S. Durkin, J. A.Eisen, C. M. Ronning, W. B. Barbazuk, M. Blanchard, C. Field, C. Halling,G. Hinkle, O. Iartchuk, H. S. Kim, C. Mackenzie, R. Madupu, N. Miller, A.Shvartsbeyn, S. A. Sullivan, M. Vaudin, R. Wiegand, and H. B. Kaplan.2006. Evolution of sensory complexity recorded in a myxobacterial genome.Proc. Natl. Acad. Sci. USA 103:15200–15205.

15. Grebe, T. W., and J. B. Stock. 1999. The histidine protein kinase superfamily.Adv. Microb. Physiol. 41:139–227.

16. Harris, B. Z., D. Kaiser, and M. Singer. 1998. The guanosine nucleotide(p)ppGpp initiates development and A-factor production in Myxococcusxanthus. Genes Dev. 12:1022–1035.

17. Higgs, P. I., K. Cho, D. E. Whitworth, L. S. Evans, and D. R. Zusman. 2005.Four unusual two-component signal transduction homologs, RedC to RedF,are necessary for timely development in Myxococcus xanthus. J. Bacteriol.187:8191–8195.

18. Hodgkin, J., and D. Kaiser. 1979. Genetics of gliding motility in Myxococcusxanthus (Myxobacterales): two gene systems control movement. Mol. Gen.Genet. 171:177–191.

19. Inouye, M., S. Inouye, and D. R. Zusman. 1979. Gene expression duringdevelopment of Myxococcus xanthus: pattern of protein synthesis. Dev. Biol.68:579–591.

20. Jakobsen, J. S., L. Jelsbak, L. Jelsbak, R. D. Welch, C. Cummings, B.Goldman, E. Stark, S. Slater, and D. Kaiser. 2004. 54 enhancer bindingproteins and Myxococcus xanthus fruiting body development. J. Bacteriol.186:4361–4368.

21. Jenal, U., and J. Malone. 2006. Mechanisms of cyclic-di-GMP signaling inbacteria. Annu. Rev. Genet. 40:385–407.

22. Julien, B., A. D. Kaiser, and A. Garza. 2000. Spatial control of cell differ-entiation in Myxococcus xanthus. Proc. Natl. Acad. Sci. USA 97:9098–9103.

23. Kaiser, D. 2004. Signaling in myxobacteria. Annu. Rev. Microbiol. 58:75–98.24. Kaiser, D. 1979. Social gliding is correlated with the presence of pili in

Myxococcus xanthus. Proc. Natl. Acad. Sci. USA 76:5952–5956.25. Kashefi, K., and P. L. Hartzell. 1995. Genetic suppression and phenotypic

masking of a Myxococcus xanthus frzF defect. Mol. Microbiol. 15:483–494.26. Kim, S. K., and D. Kaiser. 1991. C-factor has distinct aggregration and

sporulation thresholds during Myxococcus development. J. Bacteriol. 173:1722–1728.

27. Kim, S. K., and D. Kaiser. 1990. C-factor: a cell-cell signaling proteinrequired for fruiting body morphogenesis of M. xanthus. Cell 61:19–26.

28. Kimura, Y., H. Nakano, H. Terasaka, and K. Takegawa. 2001. Myxococcusxanthus mokA encodes a histidine kinase-response regulator hybrid sensor

VOL. 190, 2008 TWO-COMPONENT SYSTEMS IN M. XANTHUS 623

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

required for development and osmotic tolerance. J. Bacteriol. 183:1140–1146.

29. Kroos, L., A. Kuspa, and D. Kaiser. 1986. A global analysis of developmen-tally regulated genes in Myxococcus xanthus. Dev. Biol. 117:252–266.

30. Kruse, T., S. Lobedanz, N. M. S. Berthelsen, and L. Søgaard-Andersen.2001. C-signal: a cell surface-associated morphogen that induces and coor-dinates multicellular fruiting body morphogenesis and sporulation in M.xanthus. Mol. Microbiol. 40:156–168.

31. Kumar, A., J. C. Toledo, R. P. Patel, J. R. Lancaster, and A. J. C. Steyn. 2007.Mycobacterium tuberculosis DosS is a redox sensor and DosT is a hypoxiasensor. Proc. Natl. Acad. Sci. USA 104:11568–11573.

32. Kuspa, A., L. Plamann, and D. Kaiser. 1992. Identification of heat-stableA-factor from Myxococcus xanthus. J. Bacteriol. 174:3319–3326.

33. Kuspa, A., L. Plamann, and D. Kaiser. 1992. A-signalling and the cell densityrequirement for Myxococcus xanthus development. J. Bacteriol. 174:7360–7369.

34. Lee, B., P. I. Higgs, D. R. Zusman, and K. Cho. 2005. EspC is involved incontrolling the timing of development in Myxococcus xanthus. J. Bacteriol.187:5029–5031.

35. Leonardy, S., G. Freymark, S. Hebener, E. Ellehauge, and L. Sogaard-Andersen. 2007. Coupling of protein localization & cell movements by adynamically localized response regulator in Myxococcus xanthus. EMBO J.26:4433–4444.

36. Letunic, I., R. R. Copley, S. Schmidt, F. D. Ciccarelli, T. Doerks, J. Schultz,C. P. Ponting, and P. Bork. 2004. SMART 4.0: towards genomic data inte-gration. 32:D142–D144.

37. Li, S., B.-U. Lee, and L. J. Shimkets. 1992. csgA expression entrains Myxo-coccus xanthus development. Genes Dev. 6:401–410.

38. Lobedanz, S., and L. Søgaard-Andersen. 2003. Identification of the C-signal,a contact dependent morphogen coordinating multiple developmental re-sponses in Myxococcus xanthus. Genes Dev. 17:2151–2161.

39. Mignot, T., J. P. Merlie, and D. R. Zusman. 2005. Regulated pole-to-poleoscillations of a bacterial gliding motility protein. Science 310:855–857.

40. Mignot, T., J. W. Shaevitz, P. L. Hartzell, and D. R. Zusman. 2007. Evidencethat focal adhesion complexes power bacterial gliding motility. Science 315:853–856.

41. Mika, F., and R. Hengge. 2005. A two-component phosphotransfer networkinvolving ArcB, ArcA, and RssB coordinates synthesis and proteolysis ofsigma(S) (RpoS) in E. coli. Genes Dev. 19:2770–2781.

42. Mok, K. C., N. S. Wingreen, and B. L. Bassler. 2003. Vibrio harveyi quorumsensing: a coincidence detector for two autoinducers controls gene expres-sion. EMBO J. 22:870–881.

43. O’Connor, K. A., and D. R. Zusman. 1991. Development in Myxococcusxanthus involves differentiation into two cell types, peripheral rods andspores. J. Bacteriol. 173:3318–3333.

44. Overgaard, M., S. Wegener-Feldbrugge, and L. Søgaard-Andersen. 2006.The orphan response regulator DigR is required for synthesis of extracellularmatrix fibrils in Myxococcus xanthus. J. Bacteriol. 188:4384–4394.

45. Plamann, L., Y. Li, B. Cantwell, and J. Mayor. 1995. The Myxococcus xanthusasgA gene encodes a novel signal transduction protein required for multi-cellular development. J. Bacteriol. 177:2014–2020.

46. Pollack, J. S., and M. Singer. 2001. SdeK, a histidine kinase required forMyxococcus xanthus development. J. Bacteriol. 183:3589–3596.

47. Romling, U., and D. Amikam. 2006. Cyclic di-GMP as a second messenger.Curr. Opin. Microbiol. 9:218–228.

48. Ryjenkov, D. A., R. Simm, U. Romling, and M. Gomelsky. 2006. The PilZdomain is a receptor for the second messenger c-di-GMP: the PilZ domainprotein YcgR controls motility in enterobacteria. J. Biol. Chem. 281:30310–30314.

49. Saitou, N., and M. Nei. 1987. The neighbor-joining method: a new methodfor reconstructing phylogenetic trees. Mol. Biol. Evol. 4:406–425.

50. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: alaboratory manual, 2nd ed. Cold Spring Harbor Laboratory Press, ColdSpring Harbor, NY.

51. Schultz, J., F. Milpetz, P. Bork, and C. P. Ponting. 1998. SMART, a simplemodular architecture research tool: identification of signaling domains. Proc.Natl. Acad. Sci. USA 95:5857–5864.

52. Shi, W., and D. R. Zusman. 1993. The two motility systems of Myxococcusxanthus show different selective advantages on various surfaces. Proc. Natl.Acad. Sci. USA 90:3378–3382.

53. Singer, M., and D. Kaiser. 1995. Ectopic production of guanosine penta- andteraphosphate can initiate early developmental gene expression in Myxococ-cus xanthus. Genes Dev. 9:1633–1644.

54. Skerker, J. M., M. S. Prasol, S. P. Barrett, E. G. Biondi, and M. T. Laub.2005. Two-component signal transduction pathways regulating growth andcell cycle progression in a bacterium: a system-level analysis. PLoS Biol.3:1770–1788.

55. Søgaard-Andersen, L., F. J. Slack, H. Kimsey, and D. Kaiser. 1996. Inter-cellular C-signaling in Myxococcus xanthus involves a branched signal trans-duction pathway. Genes Dev. 10:740–754.

56. Sonnhammer, E. L., G. von Heijne, and A. Krogh. 1998. A hidden Markovmodel for predicting transmembrane helices in protein sequences. Proc. Int.Conf. Intell. Syst. Mol. Biol. 6:175–182.

57. Spratt, B. G., P. J. Hedge, S. te Heesen, A. Edelman, and J. K. Broome-Smith. 1986. Kanamycin-resistant vectors that are analogues of plasmidspUC8, pUC9, pEMBL8 and pEMBL9. Gene 41:337–342.

58. Stock, A. M., V. L. Robinson, and P. N. Goudreau. 2000. Two-componentsignal transduction. Annu. Rev. Biochem. 69:183–215.

59. Sun, H., and W. Shi. 2001. Analyses of mrp genes during Myxococcus xanthusdevelopment. J. Bacteriol. 183:6733–6739.

60. Sun, H., and W. Shi. 2001. Genetic studies of mrp, a locus essential forcellular aggregation and sporulation of Myxococcus xanthus. J. Bacteriol.183:4786–4795.

61. Takeda, S., Y. Fujisawa, M. Matsubara, H. Aiba, and T. Mizuno. 2001. Anovel feature of the multistep phosphorelay in Escherichia coli: a revisedmodel of the RcsC to YojN to RcsB signalling pathway implicated in cap-sular synthesis and swarming behaviour. Mol. Microbiol. 40:440–450.

62. Thompson, J. D., D. G. Higgins, and T. J. Gibson. 1994. CLUSTAL W:improving the sensitivity of progressive multiple sequence alignment throughsequence weighting, position-specific gap penalties and weight matrix choice.Nucleic Acids Res. 22:4673–4680.

63. Tusher, V. G., R. Tibshirani, and G. Chu. 2001. Significance analysis ofmicroarrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci.USA 98:5116–5121.

64. Uhl, M. A., and J. F. Miller. 1996. Central role of the BvgS receiver as aphosphorylated intermediate in a complex two-component regulatory phos-phorelay. J. Biol. Chem. 271:33176–33180.

65. Ulrich, L. E., E. V. Koonin, and I. B. Zhulin. 2005. One-component systemsdominate signal transduction in prokaryotes. Trends Microbiol. 13:52–56.

66. Ward, M. J., H. Lew, and D. R. Zusman. 2000. Social motility in Myxococcusxanthus requires FrzS, a protein with an extensive coiled-coil domain. Mol.Microbiol. 37:1357–1371.

67. Yang, R., S. Bartle, R. Otto, A. Stassinopoulos, M. Rogers, L. Plamann, andP. Hartzell. 2004. AglZ is a filament-forming coiled-coil protein required foradventurous gliding motility of Myxococcus xanthus. J. Bacteriol. 186:6168–6178.

68. Youderian, P., and P. L. Hartzell. 2006. Transposon insertions of magellan-4that impair social gliding motility in Myxococcus xanthus. Genetics 172:1397–1410.

624 SHI ET AL. J. BACTERIOL.

on Septem

ber 24, 2020 by guesthttp://jb.asm

.org/D

ownloaded from