Expression-based identification of genetic determinants of the bacterial symbiosis ‘Chlorochromatium aggregatum


  • Roland Wenter,

    1. Bereich Mikrobiologie, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
    Search for more papers by this author
  • Katharina Hütz,

    1. Bereich Mikrobiologie, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
    Search for more papers by this author
  • Dörte Dibbern,

    1. Bereich Mikrobiologie, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
    Search for more papers by this author
  • Tao Li,

    1. Department of Biochemistry and Molecular Biology, The Pennsylvania State University, S-234 Frear Building, University Park, PA 16802, USA.
    Search for more papers by this author
    • Present addresses: Algal Genomics Research Group, Institute of Hydrobiology, Chinese Academy of Sciences, No. 7 Donghu South Road, Wuhan 430072, China;

  • Veronika Reisinger,

    1. Bereich Botanik, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
    Search for more papers by this author
    • Center for Organelle Research (CORE), University of Stavanger, Kristine Bonnevis vei 22, N-4036 Stavanger, Norway;

  • Matthias Plöscher,

    1. Bereich Botanik, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
    Search for more papers by this author
    • TRION Pharma GmbH, Frankfurter Ring 193a, 80807 München, Germany;

  • Lutz Eichacker,

    1. Bereich Botanik, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
    Search for more papers by this author
    • Center for Organelle Research (CORE), University of Stavanger, Kristine Bonnevis vei 22, N-4036 Stavanger, Norway;

  • Brian Eddie,

    1. College of Earth, Ocean, and Environment, University of Delaware, 15 Innovation Way, Newark, DE 19711, USA.
    Search for more papers by this author
  • Thomas Hanson,

    1. College of Earth, Ocean, and Environment, University of Delaware, 15 Innovation Way, Newark, DE 19711, USA.
    Search for more papers by this author
  • Donald A. Bryant,

    1. Department of Biochemistry and Molecular Biology, The Pennsylvania State University, S-234 Frear Building, University Park, PA 16802, USA.
    Search for more papers by this author
  • Jörg Overmann

    Corresponding author
    1. Bereich Mikrobiologie, Department Biologie I, Ludwig-Maximilians-Universität München, Großhadernerstrasse 2-4, D-82152 Planegg-Martinsried, Germany.
      E-mail; Tel. (+49) 531 2616 352; Fax (+49) 531 2616 418.
    Search for more papers by this author
    • Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ), Inhoffenstraße 7B, 38124 Braunschweig

E-mail; Tel. (+49) 531 2616 352; Fax (+49) 531 2616 418.


The phototrophic consortium ‘Chlorochromatium aggregatum’ is a highly structured association of green sulfur bacterial epibionts surrounding a central, motile bacterium and is the most specific symbiosis currently known between two phylogenetically distinct bacterial species. Genes and gene products potentially involved in the symbiotic interaction were identified on the genomic, transcriptomic and proteomic level. As compared with the 11 available genomes of free-living relatives, only 186 open reading frames were found to be unique to the epibiont genome. 2-D differential gel electrophoresis (2-D DIGE) of the soluble proteomes recovered 1612 protein spots of which 54 were detected exclusively in consortia but not in pure epibiont cultures. Using mass spectrometry analyses, the 13 most intense of the 54 spots could be attributed to the epibiont. Analyses of the membrane proteins of consortia, of consortia treated with cross-linkers and of pure cultures indicated that a branched chain amino acid ABC-transporter binding protein is only expressed in the symbiotic state of the epibiont. Furthermore, analyses of chlorosomes revealed that an uncharacterized 11 kDa epibiont protein is only expressed during symbiosis. This protein may be involved in the intracellular sorting of chlorosomes. Application of a novel prokaryotic cDNA suppression subtractive hybridization technique led to identification of 14 differentially regulated genes, and comparison of the transcriptomes of symbiotic and free-living epibionts indicated that 328 genes were differentially transcribed. The three approaches were mostly complementary and thereby yielded a first inventory of 352 genes that are likely to be involved in the bacterial interaction in ‘C. aggregatum’. Notably, most of the regulated genes encoded components of central metabolic pathways whereas only very few (7.5%) of the unique ‘symbiosis genes’ turned out to be regulated under the experimental conditions tested. This pronounced regulation of central metabolic pathways may serve to fine-tune the symbiotic interaction in ‘C. aggregatum’ in response to environmental conditions.


Research on symbiotic interactions involving prokaryotes so far has focused on their associations with higher eukaryotes. However, numerous cases of highly structured, purely prokaryotic associations between phylogenetically distinct bacterial species have been documented (Overmann and Schubert, 2002). These so-called consortia include aggregates of deltaproteobacteria and archaea that mediate anaerobic methane oxidation (Boetius et al., 2000), or associations consisting of two different archaeal species (Huber et al., 2002). Monospecific cell–cell interactions have also been documented between cells of a single deltaproteobacterium species constituting the highly structured magnetotactic multicellular prokaryotes(Wenter et al., 2009). Microbial consortia are not only relevant for maintaining biogeochemical cycles in different environments, but also are of medical (e.g. in dental plaque; Whittaker et al., 1996) and technological significance (e.g. in wastewater treatment; De Bok et al., 2004). Finally, the molecular mechanisms of bacteria–bacteria interactions in consortia have implications for the evolution of bacteria interacting with human or plant hosts.

Phototrophic consortia represent the most highly developed type of bacterial interactions between different bacteria (Schink, 2002). Nineteen different types of phototrophic consortia have been described to date (Glaeser and Overmann, 2004). Depending on the type, up to 69 green sulfur bacterial epibionts surround a central colourless, rod-shaped betaproteobacterium in a highly ordered fashion. Although already known for more than a century (Lauterborn, 1906), a stable laboratory culture of the phototrophic consortium ‘Chlorochromatium aggregatum’ was established only a decade ago (Fröstl and Overmann, 1998). The epibiont of ‘C. aggregatum’ could subsequently be isolated in pure culture and was described as Chlorobium chlorochromatii CaD (Vogl et al., 2006). It is an obligately anaerobic, immotile and photolithoautotrophic green sulfur bacterium that utilizes sulfide as electron donor and thus resembles free-living green sulfur bacterial species. The epibiont cells contain chlorosomes that represent the principal photosynthetic light-harvesting structures of green sulfur bacteria. Each chlorosome consists of up to 250 000 self-assembled bacteriochlorophylls surrounded by a monogalactosyl-diglyceride monolayer and chlorosomal (Csm) proteins (Frigaard and Bryant, 2006). When in association, specific morphological adaptations are observed in the epibiont. These include an additional layered structure at the inner face of the cytoplasmatic membrane, as well as an intracellular sorting process that leaves the adhesion site void of chlorosomes (Vogl et al., 2006; Wanner et al., 2008). Intact phototrophic consortia show a chemotactic behaviour towards sulfide and 2-oxoglutarate (Fröstl and Overmann, 1998; Glaeser and Overmann, 2003). They accumulate in the light by means of a scotophobic response, whereby the central bacterium confers motility to the consortium (Fröstl and Overmann, 1998; Glaeser and Overmann, 2003). Cell division of the partner bacteria in phototrophic consortia is highly coordinated producing two complete daughter consortia (Overmann et al., 1998). These observations indicate that rapid exchange of multiple signals and highly specific reciprocal regulation mechanisms must exist between the epibionts and the central bacterium.

Recently, four putative symbiosis genes unique to the epibiont among sequenced green sulfur bacterial genomes were described. Two of the genes code for unusually large haemagglutinin-like proteins, one for a putative haemolysin and another one for a RTX toxin-like protein predicted to form a C-terminal calcium binding beta roll structure. These four genes exhibit similarity to virulence factors of typical proteobacterial pathogens and may have been transferred laterally into the epibiont genome (Vogl et al., 2008).

Chlorochromatium aggregatum’ represents the first cultivable model system available to dissect the molecular basis of the symbiotic interaction between different prokaryotes and therefore provides the unique opportunity to elucidate general principles of such highly specific interactions between different bacterial cells. In the present study we extended the analysis of the symbiosis in ‘C. aggregatum’ beyond the four known putative symbiosis genes. We combined genomic approaches with comparative transcriptomic and proteomic analyses of consortia and pure epibiont cultures in order to elucidate the genes and gene products of the green sulfur bacterial epibiont that are involved in the interaction with its non-photosynthetic partner.


Identification and classification of genes unique to the epibiont genome

The 12 available genome sequences of green sulfur bacteria cover the known phylogenetic diversity of this group (Eisen et al., 2002; Z. Liu, T. Li, F. Zhao, J. Overmann and D. A. Bryant, unpubl. results). Therefore, a comparison of the epibiont genome to the 11 genomes of the free-living relatives was used in order to identify open reading frames (ORFs) that are unique to the epibiont Chl. chlorochromatii CaD. Such genes could potentially be involved in the interaction with the central bacterium in the ‘C. aggregatum’. Pairwise comparisons by in silico subtractive hybridization analysis revealed that the epibiont genome contained 339 (compared with Chlorobium clathratiforme DSM 5477T) to 812 (compared with Chloroherpeton thalassium ATCC 35110T) unique genes. When compared with the 11 other green sulfur bacterial genomes as a whole, a total of 186 ORFs were identified to be unique for the epibiont (Table S1). Of these, 99 ORFs code for hypothetical proteins with unknown functions, 25 for proteins of DNA or protein modification (acetyl-, nucleotidyl-, glycosyl- or methyltransferases), 18 for proteins involved in cell membrane, cell wall and capsule formation, 12 for resistance against foreign DNA and phages, 6 for transcriptional regulators, another 6 for conjugative transfer, 5 for ABC-transporters, 4 for virulence factors, another 4 for signalling and regulation and 7 for proteins with other functions. This analysis extended the inventory of known epibiont ORFs with similarities to known proteobacterial virulence factors (Vogl et al., 2008). Cag_0615 codes for an outer membrane efflux protein that contains a conserved TolC-like domain and typically is part of bacterial type I secretion systems. The gene product of Cag_1408 is related to the Escherichia coli membrane fusion protein HlyD that mediates the transport of haemolysin across the periplasm as part of the type I secretion system (Balakrishnan et al., 2001), whereas the product of Cag_1570 is related to VapD, a putative toxin of a toxin–antitoxin pair found in many pathogenic bacteria (Daines et al., 2004).

Based on the information provided by the IMG database (compare Experimental procedures) and additional analyses of signal peptides and membrane protein topology using SPOCTOPUS (Viklund et al., 2008), 36 of the ORFs had a signal peptide and 28 possessed predicted transmembrane helices (Table S1). Fifteen ORFs showed both characteristics and thus most likely encode membrane proteins targeted to the cytoplasmatic membrane using the Sec machinery (Driessen and Nouwen, 2008). The unique ORFs of the epibiont genome formed 21 gene clusters with lengths between 0.98 and 22.7 kb (average length, 11 kb) corresponding to clusters of 2–23 ORFs (average, 10 ORFs).

Changes in the soluble proteome of symbiotic and non-symbiotic epibiont cells

In order to detect differentially expressed soluble proteins in the symbiotic and free-living state, a comparative proteome analysis of ‘C. aggregatum’ and axenic epibiont cultures was conducted using 2-D differential gel electrophoresis (2-D DIGE) technology (Figs 1 and S1). In order to differentiate growth phase-related from symbiosis-related effects, the protein expression patterns in the exponential and stationary phases of pure epibiont cultures were also analysed (data not shown). Three spots were detected to be present or absent depending on the growth phase and thus excluded from further analysis.

Figure 1.

Comparative 2-D DIGE analysis of the epibiont cytoplasmatic proteome in the symbiotic (red) and non-symbiotic state (green). Proteins expressed equivalently in both conditions register as yellow in the image overlay.

2-D DIGE of the soluble proteomes recovered a total of 1612 different protein spots over all gels. Extracts of consortia on average yielded 1574 spots of which 54 ± 7 (3.6%) were detected exclusively in extracts from consortia but not in pure epibiont cultures. The 13 most intense of the 54 spots could be attributed to the epibiont based on ESI-MS/MS analysis and comparison with the available genome sequence (spots labelled C1–C13; Fig. 1; Table 1), whereas the protein content of the other spots was too low to be analysed. A specific function could be assigned to 10 of the proteins. Most notably, the proteins identified as nitrogen regulatory protein P-II, 2-isopropylmalate synthase and glutamate synthase (Table 1) are likely to be involved in the central amino acid metabolism. Two proteins involved in sugar metabolism (glyceraldehyde-3-phosphate dehydrogenase, phosphotransferase protein IIA), as well as porphobilinogen deaminase and UDP-3-O-[3-hydroxymyristoyl] glucosamine N-acyltransferase were identified (Table 1). In addition to the 13 epibiont proteins, three of the proteins that were exclusively present in intact consortia were identified as phosphoenolpyruvate carboxykinase, UDP-glucose-4-epimerase and enolase (C14–C16; Fig. 1, Table 1). The respective amino acid sequences were subjected to blastx analysis ( and revealed a phylogenetic affiliation with the corresponding proteins of betaproteobacteria, thereby indicating a possible origin of these proteins from the central rod.

Table 1.  Differentially regulated proteins identified by a combination of 2-D DIGE, ESI-MS/MS and database analysis of consortia and epibiont pure cultures.
Protein descriptionProtein expression(+/− present)
Spot name/locus tagProtein nameUniProtKB/TrEMBL/EC-numbersPredicted functionCE
C1/Cag_1515Porphobilinogen deaminaseQ3AQF5/ metabolism, haem biosynthesis+
C2/Cag_1420Glyceraldehyde-3-phosphate dehydrogenaseQ3AQP7/
C3/Cag_1150Protein disulfide-isomeraseQ3ARG3/-Protein disulfide oxidoreductase activity; thioredoxin domain+
C4/Cag_1154UDP-3-O-[3-hydroxymyristoyl] glucosamine N-acyltransferaseQ3ARF9/2.3.1.-LpxD-like; Lipid A biosynthesis+
C5/Cag_1468PTS IIA domain proteinQ3AQK1/-Phosphotransferase system protein IIA with fructose-specific domain+
C6/Cag_0538Flavoprotein cofactorQ3AT64/ of glutamate synthase (Cag_0537); NAD binding+
C7/Cag_16732-Isopropylmalate synthase, yeast typeQ3AQ00/ biosynthesis+
C8/Cag_0188Putative uncharacterized proteinQ3AU59/-2-Nitropropane dioxygenase; electron carrier/oxidoreductase+
C9/Cag_0537Glutamate synthaseQ3AT65/ biosynthesis; glutamine degradation; nitrogen metabolism+
C10/Cag_1737Putative uncharacterized proteinQ3APT7/-Ferritin-like AB metal binding domain+
C11/Cag_1245Nitrogen regulatory protein P-IIQ3AR69/-Amino acid metabolism; enzyme regulator of glutamine synthetase; regulation of nitrogen utilization+
C12/Cag_0444Inorganic pyrophosphataseQ3ATF8/ hydrolysis+
C13Cag_1572Putative uncharacterized proteinQ3AQ98/-Nucleoside-diphosphate-sugar epimerase; nucleotide-sugars+
E1/Cag_0823Leucine aminopeptidaseQ3ASD6/ of an N-terminal amino acids, preferably leucine; protein processing and degradation; transcription factor+
E2/Cag_1642Oxidoreductase, short-chain dehydrogenase familyQ3AQ31/ of short-chained ketones+
C14/BB2434 Bordetella bronchiseptica RB50Phosphoenolpyruvate carboxykinaseQ7WJQ9/ metabolism via TCA cycle+
C15/NMB0064 Neisseria meningitidis MC58UDP-glucose-4-epimeraseP56985/ and nucleotide sugars metabolism+
C17/bpro_3184 Polaromonas sp. JS666EnolaseQ128E4/

In contrast to symbiotic epibionts, only two (0.1%) of all protein spots were exclusively present in the soluble proteome of the free-living epibiont (E1, E2; Fig. 1). These two proteins were identified as leucine aminopeptidase and an oxidoreductase of the short-chain dehydrogenase family (Table 1).

Further 2-D DIGE analyses were performed in order to study the effects of peptone or consortia culture supernatant on the expression of soluble proteins of the epibiont. Addition of peptone stimulated the expression of the phosphotransferase system protein IIA (Cag_1468) in pure epibiont cultures, which was otherwise only detectable in symbiotic epibionts (data not shown). However, peptone did not induce the expression of any of the 53 other differentially expressed soluble proteins. Addition of supernatant of consortia cultures caused the expression of elongation factor Tu (Cag_1853) and of phenylalanyl-tRNA synthetase (Cag_1544) (data not shown) neither of which had been detected in the symbiotic nor in the free-living state of the epibiont.

Comparison of consortia and epibiont membrane proteins

Two distinct proteins bands were found in extracts of membrane proteins of consortia when compared with those of pure epibiont cultures (Fig. 2; lanes C and E; bands CM1, CM2). Both protein bands were absent in protein extracts of the epibiont (Fig. 2A; lane E) and were missing in consortia cross-linked with 3,3′-dithiobis[sulfosuccinimidylpropionate] (DTSSP) or bis[sulfosuccinimidyl] suberate (BS3) (Fig. 2B; lanes DTSSP, BS3), indicating that the respective proteins were localized at the cell surface or in the periplasm.

Figure 2.

A. SDS-PAGE of membrane proteins of epibiont pure culture (E), of a pure culture stimulated with consortia culture supernatant (Es) and membrane proteins of consortia (C).
B. SDS-PAGE of BS3- and DTTSP-cross-linked consortia membrane proteins compared with a non-cross-linked control of consortia (C) and membrane proteins of epibiont cultures (E). Arrows indicate the presence of a filamentous haemagglutinin/type IV pilus-related protein of the central bacterium (CM1) and a most likely periplasmic binding protein of the epibiont (CM2).

In the protein band running at 40 kDa (CM2), all 10 different peptides detected by ESI-MS/MS were related to the amino acid binding protein of the branched chain amino acid ABC-transporter of Chl. chlorochromatii CaD (Cag_0853). To analyse the induction of this protein in more detail, epibiont pure cultures were grown in K3 medium supplemented with either peptone (0.05%, w/v), branched chain amino acids (leucine, valine or isoleucine in non-inhibitory concentrations of 0.1 mM or 1 mM) or consortia culture supernatant. Whereas supplementation with the former two did not affect the pattern of membrane proteins, cultivation of the epibiont with consortia culture supernatant stimulated the expression of the ABC-transporter binding protein (Fig. 2A, lane Es).

The distinct high molecular mass protein band at 100 kDa (CM1) contained different peptides related to filamentous haemagglutinin-like proteins of Burkholderia pseudomallei as well as to type IV pilus biogenesis outer membrane proteins of Delftia acidovorans and Variovorax paradoxus. Because the latter proteins have a very similar molecular weight of about 100 kDa and are phylogenetically affiliated with betaproteobacteria, the detected peptides most likely originate from the central rod. Within this band, an additional single peptide was detected, which matched a haemagglutinin-like protein of the epibiont (Cag_1053). In the epibiont genome, this gene clusters with another haemagglutinin-like protein (Cag_1055) and two haemolysin activation/secretion proteins (Cag_1054/Cag_1056). Based on the 53–65% amino acid sequence similarity to corresponding sequences of Chlorobium ferrooxidans DSM 13031T, these ORFs are not unique to the ‘C. aggregatum’ epibiont, however. Subsequent searches of the epibiont genome identified yet another filamentous haemagglutinin-like protein (Cag_1512) that shows 65% similarity to an ORF of Chl. clathratiforme DSM 5477T.

Additional experiments demonstrated that cell–cell attachment of the partner bacteria in ‘C. aggregatum’ becomes permanent if BS3 or DTSSP are added to consortia cultures. After cross-linking, consortia could not be disintegrated any more by EGTA or other Ca2+-chelating agents. These results suggest that proteins located at the cell surface are involved in the specific cell–cell binding within phototrophic consortia.

Comparison of consortia and epibiont chlorosomes

A comparison of chlorosome membrane proteins of free-living and symbiotic epibionts by SDS-PAGE revealed that a small, uncharacterized protein of 97 amino acids (aa) with a molecular mass of about 11 kDa (Cag_1285; Fig. S2, arrow; Table 1) was exclusively present in chlorosome extracts of symbiotic epibionts. Amino acid sequence analysis of the Cag_1285 gene product revealed an N-terminal signal peptide, a high proportion of C-terminal repetitive elements and the absence of membrane helices. In the genome of the epibiont Chl. chlorochromatii CaD, ORF Cag_1285 is not located near genes for any other chlorosome protein. Based on the presence of the signal peptide, this protein most likely is localized in the cell envelope and copurifies with the chlorosomes.

Phylogenetic analysis of Cag_1285 revealed a distant relationship (similarity values < 30%; Fig. 3) to putative uncharacterized proteins of the four green sulfur bacteria Chlorobium phaeovibrioides DSM 265, Chlorobium luteolum DSM273T, Chlorobium phaeobacteroides BS1 and Prosthecochloris aestuarii DSM271T as well as to proteins of human- and phytopathogenic betaproteobacteria belonging to the genera Ralstonia, Burkholderia and Neisseria (Fig. 3). No genes similar to Cag_1285 are present in the other green sulfur bacterial genomes. Most of the betaproteobacterial proteins are unidentified, but the amino acid sequence of Neisseria gonorrhoeae Q53Z52 is annotated as an outer membrane lipoprotein with unassigned function.

Figure 3.

Phylogenetic affiliation of the gene product of Chlorobium chlorochromatii Cag_1285. The phylogenetic tree was constructed using the ProteinML maximum likelihood algorithm as implemented in the ARB programme package. Bootstrap values ≥ 50% are indicated at nodes and represent percentages of 100 replicates. UniProtKB/TrEMBL accession numbers for the amino acid sequences are given in parentheses. The sequences of Polynucleobacter spp. and Herminiimomas arsenicoxidans were used as out-group. Scale bar indicates 10% substitutions per amino acid position.

Differential transcription of selected epibiont genes

Of the 15 genes that are exclusively expressed in symbiotic epibiont cells and which could be identified in the present study (13 by 2-D DIGE and 2 by cross-linking experiments), the 8 genes involved in central amino acid and sugar metabolism and in the synthesis of outer membrane, as well as the chlorosome proteins were chosen to study transcriptional regulation. RT-qPCR analyses standardized against rpoD transcripts showed that, when compared with the free-living state, the transcription of five of the genes was upregulated in consortia (Fig. 4). The most conspicuous change was observed for the gene encoding nitrogen regulatory protein P-II for which a (189 ± 10)-fold increase in transcript abundance was determined. The second largest increase of (4.4 ± 1.1)-fold was detected for protein Cag_1285, corroborating the results of our membrane protein analysis. Transcription of the genes for glyceraldehyde-3-phosphate dehydrogenase, PTS IIA domain protein and the yeast-type isopropylmalate synthase was increased slightly but significantly [between (1.5 ± 0.2)-fold and (1.5 ± 0.3)-fold; P < 0.025, t-test]. Contrary to the results of the proteome analyses, however, no increase in transcription was observed for glutamate synthase, UDP-3-hydroxymyristoyl glucosamine N-acyltransferase and the binding protein of the branched chain amino acid ABC-transporter (Fig. 4).

Figure 4.

RT-qPCR analysis of transcription upregulation of selected ORFs coding for proteins only expressed in symbiotic epibionts. *P < 0.025; **P < 0.005; ***P < 0.001.

Comparison of the transcriptomes of the symbiotic and the free-living state

Complementing the proteome-based detection of gene regulation in the epibiont, the differences in the transcriptomes of symbiotic and free-living epibiont cells were investigated using: (i) a novel prokaryotic cDNA suppression subtractive hybridization (cDNA-SSH) technique and (ii) global transcriptome analyses based on Illumina sequencing.

In the analysis by cDNA-SSH, 94 clones generated by the subtraction reactions using consortia cDNA as tester were sequenced. Ten of the clones were affiliated with epibiont genes, whereas all others contained 16S or 23S rRNA gene fragments of the central bacterium. The 10 transcripts of the epibiont originated from four different non-ribosomal genes (Table 2; Cag_1239, Cag_1244, Cag_1245 and 1246). Open reading frame Cag_1239 is absent in the genome sequences of free-living green sulfur bacteria, encodes a 2461 aa long peptide and contains nine Vibrio/Colwellia/Bradyrhizobium/Shewanella (VCBS) domain repeats. The 100 residue long VCBS domain occurs in up to 35 copies in VCBS species but in lower numbers in several other bacteria. Cag_1239 resembles its known counterparts (Yousef and Espinosa-Urgel, 2007) in that it contains three of the cadherin domains thought to participate in cell–cell adhesion. Subsequent bioinformatic searches for VCBS domains revealed that the epibiont genome encodes the three additional VCBS domain proteins Cag_0738 (8871 aa), Cag_1242 (16311 aa) and Cag_1560 (1838 aa). In contrast to Cag_1239, homologues of the latter three proteins are present in Chl. luteolum DSM 273T and Chl. ferrooxidans DSM 13031T.

Table 2.  Differentially transcribed ORFs identified by a combination of cDNA subtractive hybridization and database analysis.
Locus tagORF nameFunction predictionCE
  1. C, consortia as tester; E, epibiont pure culture as tester; +/− refers to transcripts detected/not detected.

Cag_1239VCBSVCBS and cadherin domain protein; cell–cell adhesion+
Cag_1244Nitrogenase iron proteinNitrogenase subunit NifH; nitrogen metabolism+
Cag_1245Nitrogen regulatory protein P-IIAmino acid metabolism; enzyme regulator of glutamine synthetase; regulation of nitrogen utilization+
Cag_1246Nitrogen regulatory protein P-IIAmino acid metabolism; enzyme regulator of glutamine synthetase; regulation of nitrogen utilization+
Cag_0140ATP synthase subunit A/H+-transporting two-sectorATP synthesis+
Cag_0141H+-transporting two-sector ATPase, gamma subunitATP synthesis+
Cag_0227Magnesium-chelatase, subunit HCobalamin biosynthesis protein CobN; Porphyrin and chlorophyll metabolism+
Cag_0443Pyrophosphate-energized vacuolar membrane proton pumpInorganic H+ pyrophosphatase; oxidative phosphorylation+
Cag_0474Phosphoglucomutase/phosphomannomutasePhosphomannomutase; GDP-mannose synthesis+
Cag_0475Lipid A disaccaride synthase LpxBLipid A biosynthesis+
Cag_0624ThiolperoxidasePeroxiredoxin/thioredoxin-like; post-translational modification/protein turnover+
Cag_1143Cation efflux membrane proteinCo/Zn/Cd cation transporter+
Cag_1144Putative uncharacterized proteinUnknown function+
Cag_1163FAD binding OxidoreductaseD-lactate dehydrogenase; pyruvate metabolism+

The product of Cag_1244 was identified as nitrogenase subunit NifH. The downstream ORFs Cag_1245 and Cag_1246 both encode the carbon–nitrogen regulatory protein P-II. This result is commensurate with the pronounced upregulation of transcription observed for Cag_1245 (Fig. 4) and with the detection of protein P-II by 2-D DIGE. Forty-five clones were generated from subtraction reactions using pure epibiont cultures as tester. In this experiment, 10 of the clones analysed represented epibiont mRNA transcripts, whereas the remainder contained 16S or 23S cDNA fragments of the epibiont. Sequence analysis revealed that each of the 10 mRNAs originated from genes involved in various central metabolic functions (Table 2).

A first comparative analysis of global transcriptomes of the free-living and symbiotic epibiont was achieved by Illumina sequencing of cDNA libraries prepared from a pure culture of Chl. chlorochromatii CaD and a ‘C. aggregatum’ consortium culture. After removing sequence tags that matched to rRNA, a total of 14 117 cDNA sequence tags from the epibiont grown in the consortium were mapped to the genome of Chl. chlorochromatii CaD while 33 901 sequence tags originating from the free-living cells were mapped to the genome. The decrease in the consortium likely reflects the contribution of mRNAs from the central bacterium to the cDNA library. On average, sequence tags mapped to 47% of 2002 annotated Chl. chlorochromatii CaD protein coding genes and to 76% of the 44 tRNAs. However, the coverage of individual protein coding genes was low; 51 genes had more than 10% of their length covered by reads in both samples and 102 had more than 5% of their length covered. A total of 106 genes in the symbiotic association library and 153 of the genes in the free-living library were covered at > 10%.

A total of 328 epibiont protein coding genes were determined to be significantly differentially expressed between the consortium and free-living state (Table S2). This is 16% of the 2002 annotated protein coding genes of the Chl. chlorochromatii CaD genome. Of these, 107 protein coding genes were only found in the library from ‘C. aggregatum’ consortium while 84 genes were only detected in the library from the free-living epibiont. Of the 137 genes represented in both libraries, Cag_1725, annotated as a potassium uptake protein, was most highly induced in the consortium (36-fold) while Cag_1430, annotated as a UDP-N-acetylenolpyruvoylglucosamine reductase, was most highly (10-fold) induced in the free-living epibiont. In the symbiotic state, the expression of six additional epibiont genes involved in cellular build-up and maintenance was increased by 15- to 20-fold. These genes coded for ClpX, the ATP binding subunit of Clp protease (Cag_0183), TPR repeat-containing protein (Cag_0341), DNA helicase II (Cag_1415), RecN-like DNA repair protein (Cag_1749), the 30S ribosomal protein S19 (Cag_1847) and the translation elongation factor G (Cag_1854) (Table S2). The tetratricopeptide repeat (TPR) protein mediates protein–protein interactions and the assembly of multiprotein complexes (D'Andrea and Regan, 2003). Proteins containing TPRs are involved in a variety of biological processes, such as cell cycle regulation, transcriptional control and protein folding.

Sequence tags were detected in both free-living Chl. chlorochromatii CaD and the ‘C. aggregatum’ consortium for four putative symbiosis genes identified by genomic SSH: Cag_0614, 0616, 1919 and 1920 (Vogl et al., 2008). The expression ratio indicated that mRNA for Cag_0614 (P ≤ 0.003) and Cag_1920 (P ≤ 0.009) was only 1.2–1.3-fold more abundant in consortia, while Cag_1919 decreased 1.4-fold (P ≤ 0.029), and the 1.2-fold increase for Cag_0616 was not supported as statistically significant (P ≤ 0.30). These data were consistent with end-point RT-PCR data that indicated that mRNAs of the four putative symbiosis genes were present both in the ‘C. aggregatum’ consortium and free-living Chl. chlorochromatii CaD (Vogl et al., 2008). Notably, the expression ratio for Cag_0615, located between Cag_0614 and 0616, increased 4.3-fold (P ≤ 0.002) in symbiotically associated epibiont cells relative to the free-living state. This was consistent with the prediction that an independent promoter exists for this ORF (Vogl et al., 2008).


To date, the molecular determinants of consortia formation have remained unknown. Employing the phototrophic consortium ‘C. aggregatum’ as a model system, the present comparative genomic, transcriptomic and proteomic study yielded a first inventory of genes with potential relevance for such interactions, and generates novel hypotheses regarding the molecular mechanisms that underlie the formation of morphologically and physiologically united higher entities by two or more different species, i.e. bacterial ‘heterologous multicellularity’.

Unique genes of the epibiont with functional implications for the symbiosis

Based on our in silico analysis, the genome of the ‘C. aggregatum’ epibiont contains 186 unique ORFs. While the major fraction of these ORFs code for hypothetical proteins with unknown functions and hence await future identification, the eight unique epibiont genes Cag_0614, Cag_0615, Cag_0616, Cag_1239, Cag_1408, Cag_1570, Cag_1919 and Cag_1920 match known bacterial virulence factors and therefore provide promising targets for future functional studies of the molecular coupling across the interface of the partner cells in phototrophic consortia. Of particular interest are the genes suggested to encode haemagglutinins, as the latter are typically associated with the cell membrane, are exposed at the cell surface and mediate the attachment of pathogenic bacteria to their host cells (Relman et al., 1989; Kajava et al., 2001); other homologues are involved in the adherence of non-pathogenic bacteria to surfaces and other microorganisms (Dalisay et al., 2006).

A second conspicuous group of unique genes in the epibiont genome are the seven ORFs Cag_0648 to Cag_0650, Cag_0665, Cag_0668, Cag_0673 and Cag_0675, which are closely located to each other on the genome (Table S1). Their gene products are likely to participate in capsular exopolysaccharide biosynthesis (Cag_0649, Cag_0650) or in transmembrane polysaccharide export (Cag_0648, Cag_0665). Open reading frame Cag_0668 was annotated as capA gene that is part of a minimal set of four genes (capB, C, A and E) required for poly-γ-glutamate synthesis by Gram-positive members of the Bacillales (Candela and Fouet, 2006) but is present in only 19 (2.2%) genomes of Gram-negative bacteria. Because the epibiont genome does not contain homologues of the other three cap genes, Cag_0668 and the neighbouring genes maybe involved in polysaccharide formation rather than poly-γ-glutamate synthesis. Based on electron microscopic studies, cell attachment in phototrophic consortia is mediated by a dense interconnecting network of up to 150 nm long and 3 nm wide hair-like filaments that cover the epibiont cells and in turn form an elastic capsule enclosing the central rod (Wanner et al., 2008). Furthermore, thick capsules have been observed in phototrophic consortia particularly from natural populations (Overmann et al., 1998) when compared with laboratory cultures. The synthesis of extracellular capsular material by the epibiont not only may contribute to the formation of cell aggregates but may also confer an additional protective mechanism to the bacterial cells in ‘C. aggregatum’ under suboptimal growth conditions in nature.

Preadaptation of free-living green sulfur bacteria to the symbiotic interaction

Pairwise comparisons with the genomes of 11 free-living green sulfur bacteria identified 339–812 (mean ± SD, 557 ± 122) genes to be unique to the epibiont genome. Considering the limited 16S rRNA gene sequence similarity between the 12 green sulfur bacteria analysed (90.2–97.0%; Overmann and Tuschak, 1997) these results were unexpectedly low as similar fractions of unique genes had previously been detected in much more closely (16S rRNA gene sequence similarity, 96.4%–100%) related bacterial lineages of thermophilic Synechococcus spp. (393 and 503 lineage-specific ORFs; Bhaya et al., 2007) and of two different Prochlorococcus ecotypes (364 and 923 unique genes; Rocap et al., 2003). An even higher fraction of the bacterial genome encoding niche-specific functions has been documented for enterohaemorrhagic E. coli O157:H7 that harbours 1387 genes not found in the non-pathogenic E. coli K-12 (Perna et al., 2001).

Even more conspicuous was the limited number of 186 ORFs that were found to be unique for the epibiont when compared with the 11 other green sulfur bacterial genomes as a whole. Low numbers of niche-specific genes have so far only been reported for pathogenic bacteria like Salmonella enterica (200 genes; Bowe et al., 1998), or Bacillus anthracis (141 unique proteins compared with its close relative Bacillus cereus; Read et al., 2003) and have been interpreted as indication for preadaptation of the non-pathogenic ancestor (Groisman and Ochman, 1997). We hypothesize that preadaptation of the ancestor and access to the existing gene pool of free-living green sulfur bacteria by lateral gene transfer represent two key features of the evolution of the mutualistic interaction in phototrophic consortia, and resulted in the limited number of novel symbiosis genes found in the epibiont.

The currently most promising candidates for ‘symbiosis genes’ that were already present in the free-living ancestor of the ‘C. aggregatum’ epibiont and that predisposed the ancestor to a symbiotic lifestyle are the eight non-unique epibiont genes Cag_0738, Cag_1053, Cag_1054, Cag_1055, Cag_1056, Cag_1242, Cag_1512 and Cag_1560 that matched known bacterial virulence factors and at the same time had homologues in free-living green sulfur bacteria. The product of one of these genes (the haemagglutinin-like protein of Cag_1053) cross-linked with a filamentous haemagglutinin/Type IV pilus protein of the central bacterium protein, indicating a direct role in the specific cell–cell binding to the betaproteobacterial partner. These findings also provide the first indication for virulence factor-like genes mediating multicellular adhesion in non-pathogenic bacteria.

The present study also identified the first candidate gene for the conspicuous cytological changes that occur in the epibiont in association with the central bacterium. The protein encoded by Cag_1285 copurified with the chlorosome fraction and is most likely localized in the cytoplasmic membrane or periplasmic space. It is only distantly related to its homologues in other green sulfur bacteria and was detected only in the symbiotic state. In the symbiotic state, chlorosomes are absent from the inner face of the epibiont cytoplasmic membrane at the site of attachment to the central bacterium, whereas chlorosomes are evenly distributed along the cytoplasmic membrane in epibiont cells from pure cultures (Wanner et al., 2008) as in all other green sulfur bacteria. Cag_1285 could therefore be involved in the intercellular sorting of chlorosomes in the epibiont by modifying its cytoplasmic membrane so as to exclude chlorosomes from the cell–cell adhesion site. Cag_1285 thus represents a first target for the elucidation of this peculiar cell biological process in a bacterial cell.

Future studies of the genes mentioned above and their products should include comparative functional analyses of their homologues in non-symbiotic green sulfur bacteria in order to elucidate how an existing gene inventory of free-living bacteria is modified to establish beneficial interactions with another prokaryotic species. Such analyses will also provide a guideline for the study of other types of mutualistic prokaryotic associations.

Comparison of methods to assess differential gene expression in the epibiont

The three approaches to analyse differential gene expression covered different fractions of the epibiont genome. Compared with the 2002 protein coding genes annotated in the epibiont genome (, 1520 soluble proteins were detected by 2-D DIGE in axenic epibiont cultures. Most of the additional proteins identified in the consortium proteome originate from the epibiont, because only 19% of the identified protein spots could be assigned to the central bacterium.

The transcriptome analysis covered 47% of the annotated protein genes. This limitation of the transcriptome analysis is reflected by the fact that the P-II protein (Cag_1245) was only detected by a single unique sequence tag in the consortium sample despite the 189-fold increase in transcript abundance of this ORF seen by RT-qPCR. Low coverage will disproportionately affect small ORFs like Cag_1245 (351 bp), and the chlorosome fraction associated protein Cag_1285 (291 bp) that are among the smallest proteins annotated in the ‘C. aggregatum’ epibiont genome. Deeper sequencing and improved rRNA subtraction methods will improve on this in the future. On the other hand, the transcriptome analysis proved superior to the proteomic approach in identifying regulated genes (328 vs. 17), because the quantitative comparison of protein spots in the case of the consortia proteome is inherently restricted by the interference of proteins from the central bacterium.

Combining the complementary results from analyses of the proteome (Table 1), cDNA-SSH (Table 2) and trancriptome (Table S2) the total number of genes that were differentially expressed in the symbiotic and free-living state was 352 (Fig. 5).

Figure 5.

Combined results of the proteome, cDNA-SSH and transcriptome analyses. The three approaches were largely complementary and identified 352 genes that were differentially expressed in the symbiotic and the free-living state of the epibiont. Differentially regulated genes detected by two of the approaches are stated explicitly.

The pronounced changes in epibiont gene expression involve central metabolic pathways rather than ‘symbiosis genes’

Based on our combined results, 23.3% of the genes detected in the cDNA library (328 of 1403) are differentially expressed in the symbiotic and free-living state of the epibiont. This high fraction of regulated genes is comparable to that observed in S. enterica during the switch between the extracellular and intravacuolar environment (20.6%; Eriksson et al., 2003) and to Listeria monocytogenes during temperature shock (maximum 25%; Van der veen et al., 2007), but higher than the fractions determined for many other bacteria, like for the sulfate-reducing Desulfovibrio vulgaris during a heat shock response (14%; Chhabra et al., 2006), for E. coli under 100 different randomly simulated environments (13.6%; Gianchandani et al., 2009), for Thiobacillus denitrificans switching from aerobic to denitrifying growth (10%; Beller et al., 2006) or for Thermotoga maritima switching between biofilm and planktonic populations (6%; Pysz et al., 2004).

The large fraction of differentially expressed genes notwithstanding, the genome of the epibiont encodes only a few (56) proteins for environmental sensing and regulatory responses, similar to the genomes of other green sulfur bacteria (Eisen et al., 2002). Extended analyses of role categories ( confirmed that all green sulfur bacterial genomes encompass much less (39–75) genes involved in regulation than members of other bacterial phyla (up to 764 regulatory proteins in proteobacteria). At present it remains unclear whether the pronounced changes in gene expression observed between the symbiotic and free-living state of the epibiont are controlled by the few regulatory proteins identified to date or whether the substantial fraction of regulatory proteins has been overlooked so far.

Most notably, the observed changes in gene expression of the epibiont of ‘C. aggregatum’ that were detected by the transcriptome approach largely pertain to non-unique protein coding genes (314 of the 1816 non-unique genes) rather than to the unique ‘symbiosis genes’ identified by in silico subtractive hybridization (14 of 186 ORFs, marked in red in Table S1). The majority of the regulated non-unique genes encoded components of central metabolic pathways as well as genes for replication, recombination and repair or for translation and ribosomal structure. In contrast, 15 of the 16 virulence factor-like genes were constitutively expressed based on the present and our previous (Vogl et al., 2008) study, the VCBS domain containing Cag_1239 representing the only exception. Thus, the symbiotic interaction in phototrophic consortia is comparable to the antagonistic interaction of human pathogens (Bowe et al., 1998) as it involves the differential regulation of a large number of genes of central metabolic pathways and housekeeping functions. Similar to the emergence of pathogens like B. anthracis (Read et al., 2003), an altered gene expression of such basic cellular functions may actually represent one decisive step towards the formation of the symbiosis in phototrophic consortia.

Implications for the physiological coupling between the partner bacteria in phototrophic consortia

The experimental data previously available suggested a role of 2-oxoglutarate for the physiological coupling of the partner bacteria in phototrophic consortia. Free-living green sulfur bacteria are known to excrete significant amounts of 2-oxoglutarate (Sirevag and Ormerod, 1970), which represents a potential carbon substrate of the central bacterium. Whereas intact consortia have been shown to incorporate 2-oxoglutarate (Glaeser and Overmann, 2003), Chl. chlorochromatii CaD itself does not use this compound (Vogl et al., 2006), suggesting that 2-oxoglutarate is in fact taken up by the central bacterium. Yet, the 2-oxoglutarate uptake in intact consortia appears to be controlled by the physiological state of the epibionts and does not occur in the absence of light or sulfide (Glaeser and Overmann, 2003).

Based on the results of the present study, the metabolic coupling between the partner bacteria in ‘C. aggregatum’ may also involve amino acids. Several of the differentially regulated and non-unique epibiont genes, among them 2-isopropylmalate synthase, glutamate synthase, glutamine synthetase, leucine aminopeptidase, various Nif proteins, nitrogen regulatory protein P-II and the amino acid binding protein of a branched chain amino acid ABC-transporter (Cag_0853) are involved in nitrogen metabolism. Cag_0853 exhibits high sequence similarities of 70–80% to genes present in Chl. ferrooxidans DSM 13031T and Chl. clathratiforme DSM 5477T. In other bacteria, this ABC-transporter is typically upregulated under nitrogen limiting conditions (Nikodinovic-runic et al., 2009). Together with the significantly increased expression of the nitrogenase genes nifH, nifE, nifB in the symbiotic epibiont cells, the expression pattern of Cag_0853 hence suggests that epibionts experience nitrogen limitation in the symbiotic state that maybe caused by an increased synthesis and transfer of branched chain amino acids. A transfer of amino acids or unidentified nitrogen compounds to prokaryotic epibionts has also been suggested for other interspecific bacterial interactions like the associations of Rhizobium sp. WH2K with the nitrogen fixing filamentous cyanobacterium Anabaena sp. SSM-00 (Behrens et al., 2008) and for archaeal consortia of Ignicoccus hospitalis and Nanoarchaeum equitans (Jahn et al., 2008). In the case of archaeal consortia, analyses of membrane proteins suggest that the binding protein of an ABC-transporter of I. hospitalis is involved in metabolite exchange between both partners (Burghardt et al., 2009).

Interestingly, branched chain amino acids have recently been shown to play a decisive role for the root nodule symbiosis (Prell et al., 2009). In this symbiotic system, the plant host provides branched chain amino acid and thereby causes a downregulation of their synthesis of the Rhizobium bacteroids. This in turn leads to conditional (‘symbiotic’) auxotrophy of the bacterial symbionts, a shut down of its own ammonia assimilation and the excretion of ammonia by the bacteria (Lodwig et al., 2003). The interaction in root nodules involves branched chain amino acid ABC-transporters. Based on our observations of a differential regulation of the amino acid binding protein of a branched chain amino acid ABC-transporter (Cag_0853), and of the nitrogen fixing capability of the epibiont (Vogl et al., 2006), future studies of the exchange and reciprocal control in phototrophic consortia by branched chain amino acids appear highly rewarding. Providing a first indication of such reciprocal control between the partner bacteria, the expression of the binding protein in pure epibiont cultures could also be provoked by the addition of culture supernatant from consortia, indicating that the expression of the binding protein is controlled by some sort of signal exchange with the central bacterium.

The by far most pronounced change in expression occurred for the P-II protein (Cag_1245) that was 189-fold upregulated in symbiotic epibionts. The nitrogen regulatory protein P-II interacts with various target proteins including transcription factors, key regulatory and metabolic enzymes, most of which are involved in crucial reactions in nitrogen assimilatory pathways. It responds to changes of cellular energy status and the cellular carbon–nitrogen balance in cyanobacteria (Forchhammer, 2008). Our results thus provide further evidence for nitrogen limitation of the epibiont cells in the symbiotic state. In the amino acid metabolism, P-II acts as enzyme and transcriptional regulator of the glutamine synthetase during nitrogen limitation. The fact that the transcription of glutamine synthetase (Cag_1588) was detected exclusively in consortia and parallels the upregulation of P-II confirms that P-II exerts a regulatory function also in the green sulfur bacterial epibiont.

Epibionts of phototrophic consortia seem to be specifically adapted to a symbiosis with the central rod and have never been detected free-living in nature (Glaeser and Overmann, 2004). While a comparison between a symbiotic and a free-living state was chosen in the present study in order to elicit the most pronounced response, such a switch under natural conditions is unlikely. Yet, even this most pronounced change in environmental conditions did not elicit a regulation of most of the unique symbiosis genes. Such a regulation hence seems to be dispensable for the symbiotic interaction, or the symbiotic association has been stable in nature for a sufficient time that regulation of the potential symbiosis genes no longer confers any selective advantage. Rather, the regulatory response detected in the present study pertains to the central metabolic pathways. Because a switch between non-symbiotic and symbiotic lifestyle is unlikely, regulation may serve the purpose of fine-tuning the physiological coupling between the epibiont and the central bacterium if the consortium as a whole experiences changes in environmental conditions.

Experimental procedures

Bacterial cultures and growth conditions

Cultures of the phototrophic consortium ‘C. aggregatum’ were grown in 10 l glass flasks containing anoxic K3 medium (pH 7.2) with 1 mM sulfide as electron donor and reductant (Kanzler et al., 2005). The flasks were incubated at 20°C and under continuous illumination of 20 µmol quanta m−2 s−1 incident light intensity. Under these conditions, the consortia form a dense monolayer biofilm on the inner surface of the culture vessel, which can be harvested separately from the accompanying bacteria (Pfannes et al., 2007). The pure culture of the epibiont Chl. chlorochromatii CaD was grown under the same conditions in 500 ml bottles. For induction experiments, the epibiont was grown: (i) in medium consisting of two-thirds sterile filtered supernatant of the ‘C. aggregatum’ enrichment culture plus one-third threefold concentrated K3 medium, (ii) in K3 medium containing 0.05% (w/v) peptone or (iii) in K3 medium supplemented with 0.1 mM or 1 mM each of the branched chain amino acids leucine, valine or isoleucine.

Genome sequencing

The draft genome sequence of the epibiont was determined by the Joint Genome Institute (US Department of Energy, Walnut Creek, CA) by Sanger sequencing of the ends of 12 718 small insert (∼3 kb), 10 437 medium insert (∼6 kb) and 1909 fosmid (∼36 kb) clones. The resulting sequence data were assembled using ‘phred, phrap, consed’ software package (Ewing and Green, 1998; Ewing et al., 1998; Gordon et al., 1998), which produced an initial assembly of ∼40 contigs. The contig arrangement was deduced from the information contained in paired-end reads and split genes (Yu et al., 2002), and this led to a predicted assembly containing only four scaffolds. All remaining gaps were closed by sequencing of PCR amplicons, the primers for which were designed on the basis of the predicted contig arrangement. Combinatorial PCR was performed to close the four gaps for which no scaffolding clones existed. Regions of low sequence quality were amplified by PCR and resequenced to improve and verify the sequence. The final, completed genome consisted of a single, circular DNA molecule containing 2572 079 bp (GenBank Accession NC_007514). Annotation of the resulting genomic information was performed as described (Larimer et al., 2003).

In silico subtractive hybridization analysis

In silico subtractive hybridization was conducted with the Phylogenetic Profiler available at the DOE Joint genome Institute website ( The Chl. chlorochromatii CaD genome ( was screened for single genes that had no homologues based on blastp alignments against the other 11 publicly available genome sequences of the green sulfur bacteria Chlorobaculum parvum NCIB 8327, Chlorobaculum tepidum ATCC 49652T, Chl. ferrooxidans DSM 13031T, Chlorobium limicola DSM 245T, Chl. luteolum DSM 273T, Chl. phaeobacteroides BS1, Chl. phaeobacteroides DSM 266T, Chl. phaeovibrioides DSM 265, Chl. clathratiforme DSM 5477T, Chloroherpeton thalassium ATCC 35110T and P. aestuarii DSM 271T ( A maximum e-value of 10−5 and a minimum identity of 30% were applied for identification of homologues.

Two-dimensional difference gel electrophoresis (2-D DIGE) of the soluble proteome

For the extraction of soluble proteins cells were harvested by centrifugation (at 10 000 g, 30 min and at 4°C), resuspended in 10 mM Tris-HCl buffer containing 1 mM phenylmethanesulfonyl fluoride (PMSF) and broken by five subsequent passages through a French press cell at 16 000 psi. After clarifying the crude extract by centrifugation (20 000 g, 30 min and 4°C), the supernatant was centrifuged at 200 000 g for 1 h at 4°C. Concentrations of the soluble proteins in the supernatant were measured in triplicates using the BCA Protein Assay Reagent (Thermo Scientific Pierce, Rockford, USA). Because intact consortia consist of 16 ± 3 epibiont cells per one central bacterium (biomass ratio ∼21:1) in the cultures used for proteome analyses (Wanner et al., 2008), the majority of soluble proteins in the gels were expected to originate from the epibiont. However, the potential presence of proteins of the central bacterium precluded the exact quantification of individual protein spots. Therefore we focused our analysis on only those proteins that were present in the symbiotic state of the epibiont but not detectable in axenic cultures, or vice versa. The two different fluorescently labelled protein extracts were compared directly on a single gel to avoid gel-to-gel variations.

For comparative analyses, 200 µg of soluble protein in 10 mM Tris-HCl (pH 8.5) each derived from the epibiont pure culture or the consortia culture were stained separately for 20 min at 4°C in the dark with 320 pmol of either CyDye™ Cy5 or Cy3 (DIGE Flours for Ettan DIGE, minimal dyes, GE Healthcare, Amersham, UK). The reaction was stopped by adding 1 µl of 20 mM lysine solution and incubating the resulting solution in the dark at 4°C for 10 min. After labelling, both samples were pooled, mixed with 350 µl of rehydration buffer (9 M urea, 2 M thiourea, 4% CHAPS, 1% DTT, 12.5 µM EDTA) and incubated for 30 min at room temperature. To remove insoluble protein, the samples were centrifuged briefly for 1 min at 14 000 g. Then 3.5 µl of carrier ampholyte (Merck, Darmstadt, Germany) and 1.8 µl of a 0.2% (w/v) bromophenol blue solution as tracing solution were added. A total amount of 400 µg of protein in 400 µl was immediately subjected to isoelectric focusing (IEF) on 18 cm long IPG strips spanning a linear pH gradient of 4–7 (Immobiline™ DryStrip; GE Healthcare). The IPG strips were covered with 1.5 ml of mineral oil. The voltage settings for IEF as performed with the IPphor™ II IEF system (GE Healthcare) were as follows: (i) 30 V for 10 h, (ii) a linear increase to 200 V over 20 min, (iii) 200 V for 1 h, (iv) a linear increase to 500 V over 30 min, (v) 500 V for 1 h, (vi) a linear increase to 2000 V for 2 h and (vii) 2000 V for 2 h, a linear increase to 8000 V for 2 h, followed by a hold until a total of 80 kVh was reached. After IEF the IPG strips were incubated for 15 min with gentle shaking in 10 ml of equilibration buffer (50 mM Tris-HCl pH 6.8, 6 M urea, 2% w/v SDS and 30% v/v glycerol) containing 1% (w/v) DTT in the first, and 2.5% (w/v) iodacetamide in the second step. For second dimension protein separation an Ettan Dalt II electrophoresis system (GE Healthcare) was used. IPG strips were placed on top of polyacrylamide gels (180 × 250 × 1 mm) composed of a 4.8% stacking gel containing 125 mM Tris-HCl (pH 6.8) and a 12.5% separating gel containing 375 mM Tris-HCl (pH 8.8) and 4 M urea. Ten microlitres of Cy3/Cy5-labelled wide range protein standard (ECL Plex Fluorescent Rainbow marker, GE Healthcare) was provided by a piece of filter paper (3 × 5 × 30 mm) next to the IPG strip. Gels were overlain with 0.5% (w/v) low melting agarose solution. Gels were run at 15°C with a running buffer composed of 25 mM Tris, 192 mM glycine and 0.1% (w/v) SDS. Electrophoresis was conducted at 600 V and 18 mA per gel over night and stopped when the bromophenol blue marker reached the bottom of the gel.

For each experimental condition tested, three biological replicas (independent cultures) were extracted and each culture analysed in duplicate gels (technical replicates). Overall, 24 parallel gels were thus analysed. Afterwards, the CyDye™-labelled protein gels were scanned directly between the low-fluorescence glass plates using a Typhoon Trio scanner (GE Healthcare) with a resolution of 100 µm and the photomultiplier tube set at 600 V. After scanning, the 2-D gels were stained with colloidal coomassie (Neuhoff et al., 1988) and dried on 3 mm Whatman paper (GE Healthcare) in a gel dryer. DeCyder software package (version 4.0; GE Healthcare) was used for determining the total spot numbers and for a refined control of spot boundaries. Comparative, qualitative image analysis was performed manually using ImageQuant v5.2 (Molecular Dynamics, Sunnyvale, USA). Only protein spots that were detected exclusively either in consortia extracts, in extracts of pure epibionts or in extracts of supplemented pure epibiont cultures were selected for subsequent protein identification. Spots were excised manually from dried, colloidal coomassie stained gels.

Membrane proteins

Membrane fractions of free-living epibionts were compared with those of untreated consortia in order to identify membrane proteins potentially involved in the cell–cell interactions. Parallel samples were treated with the cross-linkers BS3 and DTSSP (Thermo Scientific Pierce) that both comprise a 12 Å spacer. Because both cross-linking agents are water-soluble and cannot pass the cytoplasmic membrane, they only react with extracellular or periplasmic proteins (Luethy et al., 1995). BS3 is non-cleavable and DTSSP is thiol-cleavable. A fresh consortia biofilm was pelleted and dissolved in 5 ml of anoxic Hepes buffer (5 mM; pH 7.5) containing 5 mM BS3 or 2 mM DTSSP. After incubation for 30 min at room temperature, Tris-HCl (pH 7.5) was added to a final concentration of 50 mM and incubated for 15 min to quench the cross-linking reaction. To test the success of cross-linking epibionts and its central rod, EGTA was added to cross-linked consortia in a final concentration of 20 mM, as the cells in phototrophic consortia disaggregate in the presence of EGTA (Vogl et al., 2008).

For extraction of membrane proteins, cells were harvested and broken as described above. The homogenates were clarified by centrifugation at 20 000 g for 30 min at 4°C, and the membrane fractions were pelleted by ultra-centrifugation at 200 000 g for 1 h at 4°C. After solubilization of proteins with 2% (w/v) SDS at 45°C for 30 min, the extracts were centrifuged again at 200 000 g for 1 h and solubilized proteins in the supernatants were precipitated by adding 9 vols of acetone, followed by incubation at 0°C for 16 h. Membrane proteins were collected by centrifugation at 20 000 g for 30 min; the pellets were washed once with acetone and then resuspended in 10 mM Tris-HCl buffer containing 1 mM PMSF and 1% SDS. Protein concentrations were measured in triplicates using the BCA Protein Assay Reagent (Thermo Scientific Pierce). To break the cross-links in control assays, DTSSP-cross-linked consortia proteins were treated with 5% β-mercaptoethanol for 10 min at 56°C prior to gel electrophoresis. Forty-five micrograms of each membrane protein extract and 3 µl of protein standard (ECL Plex Fluorescent Rainbow marker, GE Healthcare) were separated by glycine-SDS-PAGE (Laemmli, 1970) in 10% (w/v) polyacrylamide gels. Electrophoresis was conducted at room temperature and 20 mA for 16 h. Gels were stained with colloidal coomassie (Neuhoff et al., 1988).

Isolation of chlorosomes and analysis of chlorosome proteins

Cultures of ‘C. aggregatum’, Chl. chlorochromatii CaD and Cba. tepidum ATCC 49652T were harvested in the exponential growth phase by centrifugation at 10 000 g and 4°C. Cells were resuspended in isolation buffer (10 mM Tris-HCl, 0.5 M betainhydrochloride, 0.5 mM PMSF and 1 mM dithiothreitol; pH 7.4), and then treated with lysozyme (3 mg ml−1) for 20 min. Cells were disrupted by two subsequent sonication steps (each for 5 min, 50% pulse and 45% power; Cell Disruptor B15, Branson, Danburry, USA) with intermittent cooling on ice (5 min). After a centrifugation at 12 000 g for 20 min, the supernatant was loaded onto a 10–60% sucrose gradient. The sucrose gradients were centrifuged at 26 000 g for 18 h. This procedure yielded a visible green band containing the chlorosomes at about 30% sucrose. The entire isolation procedure was performed in dim light and at 4°C.

Fluorescence spectroscopy was employed to assess whether chlorosomes had remained intact during the isolation procedure. The fluorescence of free bacteriochlorophyll c (BChl c) shows a maximum at 666 nm and is much more intense than the fluorescence of aggregated BChl c in intact chlorosomes that fluoresce at an emission maximum of 766 nm (Bryant et al., 2002). Therefore damages to the chlorosomes causing release of BChl c can be detected by fluorescence analysis. The chlorosome fraction was diluted 20-fold with isolation buffer, and fluorescence was measured at 450 nm (FluoroMax-3, Horiba, Kyoto, Japan). The emission spectrum showed the distinct maximum at 766 nm typical for chlorosome BChl c aggregates, indicating that chlorosomes had remained intact during the isolation process. Electron microscopy was performed as an additional control for the integrity of chlorosomes. For negative contrasting, the chlorosome-containing fractions were diluted with water in a ratio of 1:25 and dropped on a collodium-coated, carbon-steamed grid. After 1 min, the fraction was removed and the grid was dried at room temperature. Subsequently, one drop of uranyl acetate (1% w/v) was applied, the grid incubated for 1 min, dried again at room temperature and viewed in a Zeiss EM 912 electron microscope (Zeiss, Oberkochen, Germany) with an OMEGA filter in zero-loss mode. Electron microscopic inspection demonstrated that chlorosomes derived from free-living as well as from symbiotic epibionts exhibited a smooth surface and the ellipsoid shape typical for an intact chlorosome membrane.

A total of 20–100 µg of each chlorosome protein extract and 8 µl of protein marker (Polypeptide SDS-Page Standards, Bio-Rad Laboratories, Hercules, USA) were separated by Tricine-SDS-PAGE (Schägger and von Jagow, 1987) in 12% (w/v) polyacrylamide gels. Electrophoresis was conducted at 4°C and 40 mA for 20 h and afterwards the gels were stained either with colloidal coomassie (Neuhoff et al., 1988) or silver (Blum et al., 1987).

After identification of epibiont chlorosome proteins by mass spectrometry (see below), the amino acid sequence Chl. chlorochromatii CaD Cag_1285 was subjected to Expasy ‘Fasta3’ ( and blastp ( similarity searches as well as an IMG othologous cluster analysis ( The sequence of Cag_1285 and the 12 most closely related amino acid sequences recovered from the databases were imported into ARB (Ludwig et al., 2004) and aligned with the ClustalW Protein alignment tool. Maximum-likelihood trees were calculated based on 60-amino-acid positions using the ProteinML programme implemented in ARB and applying the Dayhoff amino acid substitution model matrix. The robustness of trees was inferred by bootstrap analysis after 100 resamplings using PhyML.

Mass spectrometry

Protein bands and spots were excised manually from one- or two-dimensional gels and digested with modified trypsin (Promega, Heidelberg, Germany) using OMX-S according to the protocol of the manufacturer (OMX GmbH, Wessling, Germany) (Granvogl et al., 2007). All samples were analysed by nano-LC-ESI-MS/MS on a quadrupole time-of-flight tandem mass spectrometer (Micromass ESI Q-TOF Premier, Waters, Manchester, UK). For trapping, samples were loaded onto a 5 µm Symmetry C18 180 µm × 20 mm column (Waters). Loading and washing of the sample on the trapping column was achieved at a flow rate of 5 µl min−1 for 3 min. Separation of peptide mixtures was achieved by nanoAcquity Ultra Performance Liquid Chromatography using a 1.7 µm BEH130 C18, 75 µm × 100 mm reversed phase nano column (Waters). Peptides were eluted by running a linear gradient using 100% H2O, 0.1% formic acid as solvent A and 100% acetonitrile, 0.1% formic acid as solvent B, at a flow rate of 0.4 µl min−1 within 60 min. For protein identification, sequence tags obtained from the fragment spectra were used in protein database searches. Proteins were identified by similarity searches (blast) of the amino acid sequence tags in the frame ‘fasts3’ from the European Bioinformatics Institute (EBI, The MS and MS/MS spectra were used to search the latest version of the SWISSPROT database using ProteinLynx Global Server (Waters). For manual sequence analysis of the sequence tags, the Chl. chlorochromatii CaD database was used, which was not linked to the ProteinLynx Global Server. Only proteins exactly matching predicted proteins of Chl. chlorochromatii CaD were counted as positives. False positives were excluded by checking the MS and MS/MS spectra manually.

Reverse transcription, quantitative real-time PCR (RT-qPCR)

Chlorobium chlorochromatii CaD and ‘C. aggregatum’ cultures were mixed with 12.5% ice-cold EtOH/phenol stop solution (5% phenol pH 4.5–5.5 in ethanol) to avoid RNA-degradation. Total RNA was isolated using phenol-chloroform (Chromczynski and Sacci, 1987) and subsequently purified with the RNeasy MinElute Cleanup Kit (Qiagen, Hilden, Germany) according to the protocol of the manufacturer. The RNA was treated with Turbo DNA free (Applied Biosystems, Forster City, USA) to remove all remaining DNA contamination. RNA concentrations were determined spectrophotometrically at 260 nm in a NanoDrop 1000 (Thermo Fisher Scientific NanoDrop, Wilmington, USA), and quality was assessed on a formaldehyde gel (3.1%, w/v). As a highly sensitive test for the absence of genomic DNA, a step-down PCR with a primer set targeting the rpoD gene (Cag_0387) that codes for the RNA polymerase sigma factor A of Chl. chlorochromatii CaD (Vogl et al., 2008) was conducted using 200 ng of the total RNA preparation.

Reverse transcription was performed with Superscript III Reverse Transscriptase (Invitrogen) using 400 ng of RNA and 200 ng of random hexamer primer (Eurofins MWG, Ebersberg, Germany) as recommended by the supplier. Negative controls to verify absence of contaminating genomic DNA were prepared by omitting the reverse transcriptase. RT-qPCR reactions with custom designed gene-specific primers were run at optimized PCR conditions (Table S3) in an iQ5 real-time PCR detection system (Bio-Rad) using 12.5 µl of iQ SYBR Green Supermix (Bio-Rad), 10 ng of cDNA and 160 nM of each primer in a final volume of 25 µl. Tenfold serial dilutions of the cDNA were used to construct standard curves for PCR efficiency determination. Transcript quantities of each target gene were normalized to the transcript quantities of the housekeeping gene rpoD (primers rpoDf and rpoDr; Table S3). Changes in the relative abundance of transcripts between the symbiotic and free-living state of the epibiont were determined using the Pfaffl method (Pfaffl, 2001).

Prokaryotic cDNA suppression subtractive hybridization

Total RNA was isolated and DNase-treated as described above. The separation of the mRNA from 40 µg of total RNA was performed with the MICROBExpress bacterial mRNA enrichment kit (Applied Biosystems) according to the instructions of the manufacturer. Multiple preparations were pooled and concentrated with the RNeasy MinElute Cleanup Kit (Qiagen). The efficiency of rRNA depletion in the mRNA extract was analysed using an Agilent 2100 bioanalyser with a RNA LabChip (Agilent Technologies, Santa Clara, USA).

First-strand cDNA synthesis was done with Superscript III Reverse Transscriptase (Invitrogen) as described by the manufacturer, using 2 µg of mRNA 0.5 µl of 10 µM PCS primer (De long et al., 2008) and 400 U of reverse transcriptase. After 90 min of incubation, additional 400 U of the enzyme were added and the incubation continued for another 90 min. Two first-strand synthesis reactions were pooled and the second strand cDNA synthesis performed with the PCR-Select™ cDNA substraction kit (Clontech Laboratories, Mountain View, USA) according to the manufacturers instructions. One microlitre of DNase-free RNase (500 µg ml−1) (AppliChem, Darmstadt, Germany) was added and the samples incubated for 30 min at 37°C. The cDNA was purified with a QIAquick PCR purification kit (Qiagen) and yields were quantified by absorbance at 260 nm. Tester and driver were digested with RsaI and purified using the MinElute reaction cleanup kit (Qiagen). Successful digestion of the cDNA was verified by gel electrophoresis on a 1% agarose gel stained with ethidium bromide. Adaptor ligation and the ligation efficency test using the primer pair rpoDf/rpoDr (Table S3) were performed according to the manual for the PCR-Select™ cDNA substraction kit (Clontech). The first and second hybridizations were done following the instructions of the kit with the exception that 3 µl of denaturated driver and 1 µl of 4× hybridization buffer were used during the second hybridization. Primary and secondary nested suppression PCRs were run in a GeneAmp 9700 Thermal Cycler (Applied Biosystems) with Advantage 2 cDNA polymerase mix (Clontech) according to the PCR-Select™ cDNA substraction kit (Clontech) protocol using 50 nM of nested-PCR primer 2R in the secondary PCR. Amplification products of this secondary PCR were cloned using the TOPO TA cloning kit (Invitrogen). Plasmids were isolated from selected clones and inserts were sequenced with M13 forward and reverse primers (Invitrogen). Sequence analysis was performed on basis of the annotated genome sequence of Chl. chlorochromatii. The complete ORFs of the gene fragments of the DNA sequences obtained by cDNA subtractive hybridization were obtained from the DOE Joint genome Institute website (

Illumina cDNA sequencing

Aliquots of the cDNA prepared for cDNA-SSH (see above) were used for a comparison of the transcriptome of symbiotic with free-living epibionts. Prior to library construction, cDNA was fragmented using the Applied Biosystems RNA fragmentation reagents according to the instructions of the manufacturer. Libraries for sequencing were then prepared using the Illumina mRNA-Seq kit beginning with end repair of the cDNA fragments prior to adapter ligation. Sequencing was performed at the University of Delaware using the Illumina Genome Analyzer II. The loading volume for cluster generation was optimized using a qPCR assay to determine the concentration of properly constructed fragments (B. Kingham, pers. comm.).

Raw sequence data were trimmed to 30 bp and mapped to the Chl. chlorochromatii CaD genome (NC_007514) in the Eland software package with a mismatch tolerance of ≤ 2 bp. Coverage of annotated protein coding genes (reference database NC_007514.ptt), RNA coding genes (reference database NC_007514.rnt) and intergenic regions > 50 bp was determined using custom Perl scripts (available from T. Hanson on request) to calculate the number of sequence tags that mapped to each genomic region. The library size for symbiotic epibiont cells was 14 117 tags and that for free-living epibionts 33 901 tags. Accordingly, the expression ratio of genes was calculated as: (tags per gene in consortium library/14 117)/(tags in free-living library/33 901). Differentially expressed genomic regions were defined as those with an expression ratio of ≤ 0.67 or ≥ 1.5 and resulting in a P-value of < 0.01 when calculated as described by Audic and Claverie (1997) using the number of non-ribosomal matches to the genome as the library size for each sample. For sequences present in only one library, the P-value criterion ensured that at least four sequence tags matching that gene were observed to include it in the differentially expressed list.


The skilful help of Dr Bernhard Granvogl with MS data analysis is gratefully acknowledged. J. Overmann and D.A. Bryant gratefully acknowledge the contributions of the Joint Genome Institute in the draft sequencing and genome annotations for nine green sulfur bacteria. This work was supported by grants from the US National Science Foundation (MCB-0523100) and Department of Energy (DE-FG02-94ER20137) to D.A. Bryant and grants of the Deutsche Forschungsgemeinschaft to J. Overmann (DFG OV20/10-1 and 10-2).