Gamma- and epsilonproteobacterial ectosymbionts of a shallow-water marine worm are related to deep-sea hydrothermal vent ectosymbionts


E-mail; Tel. (+49) 421 2028 932; Fax (+49) 421 2028 580.


The marine oligochaete worm Tubificoides benedii is often found in high numbers in eutrophic coastal sediments with low oxygen and high sulfide concentrations. A dense biofilm of filamentous bacteria on the worm's tail end were morphologically described over 20 years ago, but no further studies of these epibiotic associations were done. In this study, we used fluorescence in situ hybridization and comparative sequence analysis of 16S rRNA and protein-coding genes to characterize the microbial community of the worm's tail ends. The presence of genes involved in chemoautotrophy (cbbL and cbbM) and sulfur metabolism (aprA) indicated the potential of the T. benedii microbial community for chemosynthesis. Two filamentous ectosymbionts were specific to the worm's tail ends: one belonged to the Leucothrix mucor clade within the Gammaproteobacteria and the other to the Thiovulgaceae within the Epsilonproteobacteria. Both T. benedii ectosymbionts belonged to clades that consisted almost exclusively of bacteria associated with invertebrates from deep-sea hydrothermal vents. Such close relationships between symbionts from shallow-water and deep-sea hosts that are not closely related to each other are unusual, and indicate that biogeography and host affiliation did not play a role in these associations. Instead, similarities between the dynamic environments of vents and organic-rich mudflats with their strong fluctuations in reductants and oxidants may have been the driving force behind the establishment and evolution of these symbioses.


Until recently, the epibiotic microflora of animals was often thought to be just a consortium of casual associates, but it is now becoming increasingly clear that it consists of a highly specialized community of microbes that form specific and stable symbioses with their hosts (McFall-Ngai, 2008). In chemosynthetic environments such as hydrothermal vents or sulfide-rich sediments, some marine animals are regularly colonized by ectosymbiotic bacteria (Dubilier et al., 2008; Goffredi, 2010). In coastal sediments these include nematode worms, amphipods and colonial ciliates all of which generally host only a single sulfur-oxidizing gammaproteobacterial symbiont (Gillan and Dubilier, 2004; Bayer et al., 2009; Rinke et al., 2009). Hosts from deep-sea hydrothermal vents such as shrimps, polychaete worms, crabs and barnacles often have a more diverse epibiotic community that is dominated by Gamma- and Epsilonproteobacteria (Goffredi et al., 2008; Suzuki et al., 2009; Watsuji and Takai, 2009; Petersen et al., 2010). The role of these epibiotic bacteria is generally not well understood. They have been hypothesized to contribute to their hosts' nutrition (Rieley et al., 1999; Rinke et al., 2006; Suzuki et al., 2009), detoxify sulfide and heavy metals (Alayse-Danet et al., 1987; Prieur et al., 1990), or provide defense against pathogenic microbes and predators (Gil-Turnes and Fenical, 1992; Goffredi et al., 2004a).

The tubificid oligochaete Tubificoides benedii (d'Udekem, 1855) is a marine worm that is commonly found in coastal mudflats of the North Atlantic, especially those with a high input of organic matter (Timm and Erséus, 2009). As other tubificids (Guérin and Giani, 1996), T. benedii feeds head down in the sediment and uses its tail end for respiration by holding it above the sediment in the oxygenated seawater and moving it in a swaying motion (Dubilier et al., 1995a). The worms can tolerate extended periods of low oxygen concentrations or anoxia by switching to an anaerobic metabolism, a strategy that is also used when sulfide concentrations become too high (Dubilier et al., 1994).

Tubificoides benedii from mudflats in the Wadden Sea off the coast of Germany are covered with a morphologically diverse assemblage of ectobacteria inthe mucus layer covering their body wall (Giere and Rhode, 1987). Their tail ends are regularly colonized by a dense film of filamentous bacteria. Some of these penetrate the body wall and the basal cells are anchored in the cuticle just above the worm's epidermis, suggesting a highly specific association (Dubilier, 1986). Seasonal ‘blooms’ of the filamentous bacteria have been observed, with the densest abundance occurring in the summer and fall months when sulfide concentrations are high in the worm's habitat (Dubilier, 1986). This observation together with the morphological similarity of the filamentous bacteria to sulfur-oxidizing bacteria such as Thiothrix lead to the hypothesis that these ectofilaments might be chemoautotrophic sulfur oxidizers (Dubilier, 1986), but no further studies on this association followed. The purpose of this study was therefore to characterize the phylogeny and metabolic potential of the filamentous ectobionts of T. benedii using fluorescence in situ hybridization (FISH) and comparative sequence analysis of 16S rRNA genes as well as genes involved in chemoautotrophy (cbbL, cbbM and aclB) and sulfur metabolism (aprA and soxB).

Results and discussion

General bacterial diversity

Comparative 16S rRNA gene analyses and FISH of T. benedii tail ends revealed a diverse bacterial community. The filamentous ectosymbionts were identified as Gamma- and Epsilonproteobacteria (see below). As this study focuses on these ectosymbionts, all other bacteria associated with this worm are only briefly described here.

Gammaproteobacterial sequences dominated the 16S rRNA clone libraries (52.4%), Deltaprotobacteria (20.5%) and Cytophaga/Flavobacterium/Bacteroides (CFB; 19.2%) were also abundant, and 4% of the sequences belonged to the Epsilonproteobacteria (Table 1 and Figs S1 and S2). FISH analyses with group specific probes for Deltaproteobacteria and CFB (Table 2) confirmed that bacteria from these phyla colonized the mucus layer covering T. benedii tail ends (Fig. S3). Sequences found in only low abundance in single individuals made up only 2.7% of all 16S rRNA clone library sequences and belonged to the Acidobacteria, Verrucomicrobia, Planctomycetes and candidate Division OD1 (Table 1).

Table 1.  Clone library sequences from T. benedii tail ends.
Gene/proteinClone groupAccession number% in clone libraryClosest relativeIdentity BLAST (%)Closest cultured relativeIdentity BLAST (%)
  1. Four worms were prepared and analysed individually for the 16S rRNA, three for the cbbL, cbbM and aprA clone libraries (numbers in the first column show the total number of clones sequenced for each gene). Sequences were grouped together if they had at least 99.0% nucleotide identity (16S rRNA gene) or 95% amino acid identity (cbbL, cbbM and aprA genes). Only 16S rRNA gene sequences found in at least four clones and two individuals are shown (sequences at lower abundances are grouped under ‘other’).

  2. n.a., not applicable.

16S rRNAGamma 1 ectosymbiontGU19743412.8Shinkaia crosnieri setae associated clone, AB44017595Leucothrix mucor, X8727790
  Gamma 2GU19744122.6Sulfide-microbial incubator clone, DQ22865797Dechloromarinus chlorophilus, AF17035990
n = 298 Gamma 3GU1974251.7Urothoe poseidonis ectosymbiont, AY42661395Thiothrix eikelboomii, AB04254191
  Gamma 4GU1974451.3Solemya velum endosymbiont, M9041594Thiobacillus prosperus, EU65329191
  Gamma 5GU19742212.0Environmental clones, e.g. hydrothermal vent, AB25242797n.a. 
  Other Gamma 3.4    
 Epsilon 1 ectosymbiontGU1974361.3Rimicaris exoculata Epsilon 3 ectosymbionts, e.g. FM20339597Sulfurovum lithotrophicum, AB09129294
  Epsilon 2GU1974481.7Environmental clones, e.g. methane seep, FJ26467699Sulfurovum lithotrophicum, AB09129294
  Other Epsilon 1.0    
  Delta 20.5    
  CFB 19.2    
  Other 7.1    
CbbLCbbL 1aGU19748254.0Solemya velum endosymbiont, AAT0142998Mariprofundus ferrooxydans, EAU5385289
 CbbL 1bGU19748110.3Solemya velum endosymbiont, AAT01429 (prot)96Mariprofundus ferrooxydans, EAU5385290
n = 165CbbL 1cGU19747624.8Solemya velum endosymbiont, AAT0142993Mariprofundus ferrooxydans, EAU5385289
 CbbL 2GU1974794.8−> Mariprofundus ferrooxydans, EAU5385294
 CbbL 3GU1974783.6−> Ralstonia metallidurans, ABF0838492
 CbbL 4GU1974802.4−> Ralstonia metallidurans, ABF0838497
 CbbL 5GU1974770.6−> Thiobacillus denitrificans, AAB7069796
CbbMCbbM 1GU1973942.3Lamellibrachia sp. endosymbiont 2, BAA9443393Thiobacillus denitrificans, AAA9917885
 CbbM 2GU1973984.0Lamellibrachia sp. endosymbiont, CAQ6347595Halothiobacillus neapolitanus, AAD0244291
n = 175CbbM 3aGU19739690.3Lamellibrachia sp. tube associated clone, CAQ6347994Thiomicrospira crunogena, DQ27253586
 CbbM 3bGU1973972.3Lamellibrachia sp. tube associated clone, CAQ6347990Thiomicrospira halophila, ABD5228387
 cbbM 4GU1973951.1−> Rhodopseudomonas palustris, ABD8844386
AprAAprA I1GU1974000.4O. ilvae Gamma 3 symbiont, CAJ8124596Thiobacillus thioparus, ABV8009294
 AprA II1GU1973990.4Ifremeria nautilei symbiont, ABV8004595Thiothrix nivea, ABV8002592
n = 242AprA II2GU19740132.5Lamellibrachia sp. endosymbiont, CAQ6349291Thiothrix nivea, ABV8002590
 AprA II3GU1974060.8Sea urchin gut clone, CAT0360396Thiocapsa pendens, ABV80001589
 AprA II4GU19740460.1Oligobrachia haakonmosbiensis endosymbiont, CAP0314390Thiobacillus thioparus, ABV8002787
 AprA II5GU1974031.2Lucinoma aff. kazani symbiont, CAJ8565394Thiobacillus plumbophilus, ABV8002191
 AprA II6GU1974020.4Lucinoma aff. kazani symbiont, CAJ8565390Thiobacillus denitrificans, ABV8003188
 AprA SRB1GU1974054.1−> Desulfotomaculum thermoacetoxidans, ABR9258878
Table 2.  Oligonucleotide probes used in this study (probes specific to the Gamma 1 and Epsilon 1 ectosymbionts in bold). Additional probe details are available at probeBase (Loy et al., 2007).
ProbeSpecificityProbe sequence (5′-3′)PositionaFA [%]bReference organism
  • a. 

    Position in the 16S rRNA of Escherichia coli.

  • b. 

    Formamide concentrations used in the FISH and CARD FISH hybridization buffer in % (v/v).

  • c. 

    Position in the 23S rRNA of E. coli.

  • nd, not determined.

EUBI-IIIEubacteriaGCW GCC WCC CGT AGG WGT338–35540(Daims et al., 1999)
NON338AntisenseACT CCT ACG GGA GGC AGC338–35540(Wallner et al., 1993)
GAM42aGammaproteobacteriaGCC TTC CCA CAT CGT TT1027–1043c30–40(Manz et al., 1992)
EP404T. benedii Epsilon 1 ectosymbiont and Epsilon 2
Epsilonproteobacteria (most Campylobacterales), various uncultured bacteria including associated bacteria of C. squamiferum, R. exoculata ectosymbionts, R. pachytila
AAA KGY GTC ATC CTC CA404–42040(Macalady et al., 2006)
TbGAM1-138T. benedii Gamma 1 ectosymbiont
associated bacteria of V. osheai, AB239761, S. crosnieri, e.g. AB440174, K. hirsuta, EU265799, vent clone AB464819
GGC TTG TCC CCC ACT ACT138–15540Ectothiorhodospira mongolicum, other T. benedii epibionts
TbGAM1-444T. benedii Gamma 1ectosymbiont
associated bacteria of S. crosnieri, e.g. AB440175, K. hirsuta, EU265799, some environment clones e.g. AM778459, FJ169979
CTT AAC CCC TTC CTC ACA444–46140n.d.
TbGAM2-447T. benedii Gamma 2, invertebrate burrow clones, FJ753075 and FJ753097AAG CTT AGG CTT TTC GTC447–46440n.d.
TbEP1-744T. benedii Epsilon 1
Epsilonproteobacteria, most Sulfurovum, including invertebrate associated clade, C. squamiferum, V. osheai, A. pompejana and R. exoculata Epsilon 1–5 ectosymbionts, few unculitvated, e.g. AF407203
TCT CAG CGT CAG TAC TGT744–76135n.d.
TbEP1-94T. benedii Epsilon 1 ectosymbiont
associated bacteria of S. crosnieri, e.g. AB440164, Rimicaris sp., FM203397-99, environ. clones most deep sea, e.g. FJ905659
CCG TTC GCC ACT CGA CAG95–9940T. benedii Gamma 1
CF319aMost Flavobacteria, some Bacteroidetes, some SphingobacteriaTGG TCC GTG TCT CAG TAC319–33640(Manz et al., 1996)
Most Deltaproteobacteria and most GemmatimonadetesAGT TAG CCG GTG CTT CCT
495–51240(Loy et al., 2002)

Phylogenetic analyses revealed that many of the gamma- and epsilonproteobacterial sequences from the T. benedii 16S rRNA clone libraries were most closely related to symbiotic and free-living bacteria from chemosynthetic environments (Fig. 1). Specific FISH probes were developed for all dominant gamma- and epsilonproteobacterial phylotypes (Tables 2 and S1). Aside from the filamentous ectosymbionts described below, the only other gamma- or epsilonproteobacterial phylotype from the 16S rRNA clone libraries that could be found with FISH was a rod- to cocci-shaped bacterium named the Gamma 2 ectobiont (Fig. S3). This phylotype was highly abundant in the 16S rRNA clone libraries (23% of all clones, Table 1), but with FISH, the Gamma 2 ectobiont was only observed occasionally in the mucus membrane of the worm's tail end. Phylogenetic analyses placed this ectobiont in a clade that included the Gamma 3 endosymbionts of the gutless oligochaete worms Olavius ilvae and Olavius algarvensis from Mediterranean seagrass sediments (Fig. 1) (Ruehland et al., 2008). Such a close phylogenetic relationship between ecto- and endosymbionts of hosts that are separated by large geographic distances, live in different habitats, and are not closely related to each other have rarely been described but are not unprecedented. They have, for example, also been observed between the ectosymbionts of nematodes and the Gamma 1 endosymbionts of gutless oligochaetes (Bayer et al., 2009).

Figure 1.

Phylogenetic placement of Gamma- and Epsilonproteobacteria associated with T. benedii tail ends. Consensus tree based on maximum likelihood (ML) analyses of 16S rRNA sequences with nodes that differed in other treeing analyses collapsed to a consensus branch. Sequences from this study in red, in parentheses the number of sequences with > 99.0% to the given sequence. Bacterial sequences from the clone libraries that were not found with FISH on the worms were named ‘T. benedii associated’. Five deltaproteobacterial sequences were used as an outgroup (arrow). Scale bar = 0.10 estimated substitutions per site. Numbers next to nodes correspond to bootstrap values based on 1000 ML replicates (only values above 60% are shown).

Morphology and phylogeny of the filamentous ectosymbionts

Transmission electron microscopy (TEM) showed that two filamentous morphotypes co-occurred on T. benedii tail ends (Fig. 2). Thinner filaments of 0.4–0.65 µm penetrated the worm's cuticle, with the basal cells embedded within the cuticle just above the host's epidermis (Fig. 2B–D). Vesicle-like structures were regularly observed between the basal cell of these filaments and the worm's epidermal cells (Fig. 2D), indicating interactions between the filaments and host tissues (Dubilier, 1986). Thicker filaments of 0.9–1.1 µm were attached to the cuticle but were never observed penetrating it (Fig. 2B). FISH with probes specific to the Gamma 1 and Epsilon 1 sequences shown in Fig. 1 revealed that these originated from two filamentous morphotypes on the worm's tail (Fig. 2B–D). The diameters of these two filament types in FISH micrographs corresponded to those measured with TEM in the thinner and thicker morphotypes (0.4–0.7 µm for the Epsilon 1 and 0.7–1.3 µm for the Gamma 1). This indicates that the thinner filaments observed with TEM in the worm's cuticle belonged to the Epsilon 1 ectosymbionts. This conclusion is supported by FISH analyses that show the Epsilon 1 ectosymbionts embedded within the cuticle (Fig. 2H). The distribution and abundance of the two filamentous ectosymbionts varied considerably within and between individual worms. In some cross-sections only one filament type was observed, in others, especially towards the posterior end of the tail, both the Gamma 1 and Epsilon 1 ectosymbionts co-occurred in thick patches (Fig. 2E). Overall, the abundance of both ectosymbionts was equal based on FISH analyses. The much lower abundance of Epsilon 1 sequences in the 16S rRNA clone libraries could have been caused by differences in the extraction efficiency of their DNA, primer mismatch and/or polymerase chain reaction (PCR) bias, as described from many other studies of mixed microbial communities (Gonzalez and Moran, 1997; Kanagawa, 2003; Sipos et al., 2007; Hong et al., 2009).

Figure 2.

Microbial community of T. benedii tail ends.
A. Light microscopy showing the numerous filaments on the worm's tail.
B–D. Transmission electron microscopy. Only the thinner filaments penetrate the cuticle (B, C, D), while the thicker filaments (B) are attached to the surface of the worm. D. Vesicle-like structure between the worm's epidermis and the basal cell of a thin filament.
E–H. Epifluorescence microscopy using FISH (E) and CARD-FISH (F–H). The thinner Epsilon 1 ectosymbionts are shown in green with the general epsilonproteobacterial probe EP404 (E) and the specific Epsilon 1 symbiont probe TbEP1-94 (F–H). The thicker Gamma 1 ectosymbionts are shown in red with the specific Gamma 1 symbiont probe TbGAM1-138 (E–H).

Phylogenetic analyses revealed that the Epsilon 1 ectosymbiont belonged to the newly established family Thiovulgaceae within the Epsilonproteobacteria (Campbell et al., 2006) and the Gamma 1 ectosymbiont to the Leucotrix/Thiothrix group within the Gammaproteobacteria (Fig. 1). Both T. benedii ectosymbionts belonged to clades consisting nearly exclusively of bacteria associated with invertebrate hosts from deep-sea hydrothermal vents (γ- and ε-ectobiont clades in Fig. 1). These hosts included the galatheid crab Shinkaia crosnieri (Watsuji and Takai, 2009), the Yeti crab Kiwa hirsuta (Goffredi et al., 2008), the scaly foot snail Chrysomallon squamiferum (Goffredi et al., 2004b), the stalked barnacle Vulcanolepas osheai (Suzuki et al., 2009), and the shrimp Rimicaris exoculata (Petersen et al., 2010). Filamentous ectobionts have been described from all of these hosts but a clear assignment of sequences from the γ- and ε-ectobiont clades to bacteria with a filamentous morphology has only been shown for R. exoculata (Petersen et al., 2010) and V. osheai (Suzuki et al., 2009). Free-living filamentous bacteria of the genus Thiothrix fall in a more distantly related but neighbouring clade to the T. benedii Gamma 1 ectosymbionts (Fig. 1), and within this clade filamentous ectosymbionts have been identified from two other animal hosts, the marine amphipod Urothoe poseidonis (Gillan and Dubilier, 2004) and the freshwater cave amphipod Niphargus ictus (Dattagupta et al., 2009).

The closest cultured relative of the T. benedii Gamma 1 ectosymbiont is the filamentous bacterium Leucothrix mucor, while the Epsilon 1 ectosymbiont is most closely related to the coccoid- to oval-shaped Sulfurovum lithotrophicum (Table 1, Fig. 1). Both L. mucor and S. lithotrophicum are sulfur oxidizers (Grabovich et al., 1999; Inagaki et al., 2004) and there is evidence for autotrophy and the use of reduced sulfur compounds as energy sources by the filamentous ectobionts of R. exoculata (Polz et al., 1998) and N. ictus (Dattagupta et al., 2009). The close relationship of the T. benedii ectosymbionts to symbiotic and free-living chemosynthetic bacteria and the sulfidic environment of the worms therefore led us to examine the potential of the T. benedii association to gain its energy through chemoautotrophic sulfur oxidation.

Metabolic potential for chemoautotrophy and sulfur oxidation

The metabolic potential of the T. benedii bacterial community was assessed by analysing protein-coding genes for sulfur metabolism (aprA and soxB) and chemoautotrophy (cbbl and aclB) (Table 1). The aprA gene, coding for the alpha-subunit of adenosine-5′-phosphosulfate (APS) reductase, is widespread in sulfur-oxidizing microorganisms including many Gammaproteobacteria, but has not yet been found in Epsilonproteobacteria (Meyer and Kuever, 2007). The majority of AprA sequences from the T. benedii microbial community (95% or 231 clones) belonged to the AprA lineage II of sulfur-oxidizig bacteria that includes sequences from both symbiotic and free-living Gammaproteobacteria (Meyer and Kuever, 2007). Two sequences were highly abundant within this lineage (T. benedii AprA II2 and II4) and both were closely related to AprA sequences from the sulfur-oxidizing gammaproteobacterial endosymbionts of tubeworms from cold seeps (Table 1, Fig. 3). Given the absence of the aprA gene from Epsilonproteobacteria, it is most likely that the AprA lineage II sequences from T. benedii originated from their gammaproteobacterial ectobionts, thus indicating their potential for the use of reduced sulfur compounds as an energy source. However, because of the incongruence of AprA phylogeny with 16S rRNA phylogeny, it is not clear which AprA sequence(s) might have originated from the T. benedii Gamma 1 ectosymbiont.

Figure 3.

AprA phylogeny based on ML analyses of deduced amino acid sequences of the aprA gene. Sequences from this study in red, in parentheses the number of sequences with > 95.0% to the given sequence. Scale bar = 0.10 estimated substitutions per site. Numbers next to nodes show bootstrap values based on 100 ML replicates (only values above 60% are shown). SOX, sulfur-oxidizing bacteria; SRB, sulfate-reducing bacteria; SRA, sulfate-reducing archaea.

A small number of T. benedii AprA sequences (4%) belonged to the lineage of AprA sequences from sulfate-reducing bacteria (Fig. 3). In sulfate reducers, the enzyme functions in the opposite direction as in sulfur oxidizers (Meyer and Kuever, 2007). The presence of AprA sequences in T. benedii from the sulfate reducer lineage corresponds well with our 16S rRNA results showing that deltaproteobacterial sequences related to sulfate-reducing bacteria were abundant in the clone libraries (20%, see Table 1 and Fig. S2). Sulfate-reducing bacteria are well known as endosymbionts in gutless marine oligochaetes where they co-occur with sulfur-oxidizing endosymbionts and engage in syntrophic cycling of reduced and oxidized sulfur compounds (Dubilier et al., 2001; Woyke et al., 2006). Evidence for co-occurring sulfur-oxidizing and sulfate-reducing bacteria has also been described for the ectosymbiotic community of the Yeti crab K. hirsuta (Goffredi et al., 2008).

To determine the potential of the Epsilon 1 ectosymbiont for sulfur oxidation, we attempted to amplify the soxB gene, coding for the SoxB component of the Sox enzyme complex. This gene is widespread among sulfur-oxidizing bacteria and has been found in all epsilonproteobacterial and some gammaproteobacterial sulfur oxidizers (Meyer et al., 2007; Ghosh et al., 2009). We were not able to amplify this gene from DNA extracted from T. benedii tail ends, despite the use of degenerate primers that work well for a phylogenetically wide range of bacteria (Petri et al., 2001; Meyer et al., 2007). It is possible that the concentrations of Epsilon 1 ectosymbiont DNA were too low for successful amplification. This symbiont was underrepresented in the 16S rRNA clone libraries (1.3%, Table 1), despite our FISH studies showing its high abundance (Fig. 2E,H).

We assessed the potential of the T. benedii microbial community for autotrophic carbon fixation by analysing key genes of the Calvin Benson Bassham (CBB) cycle (cbbL and cbbM) and the reductive tricarboxylic acid (rTCA) cycle (aclB). The cbbL and cbbM genes, coding for the large subunits of the form I and II ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO), respectively, are widespread among autotrophic organisms including many Gammaproteobacteria, but have not been found in Epsilonproteobacteria. Both genes could be amplified in DNA extracted from T. benedii tail ends, indicating the potential of its microbial community for autotrophy. The dominant sequences in the cbbL clone libraries were most closely related to sequences from the chemoautotrophic sulfur-oxidizing endosymbiont of the clam Solemya velum, while for cbbM, the dominant sequences grouped with sequences from bacteria associated with the seep tubeworm Lamellibrachia sp. (Fig. 4, Table 1). As with the aprA gene, given the many inconsistencies between phylogenetic trees based on cbbL and cbbM genes versus the 16S rRNA gene, it is not possible to identify which of these genes might have originated from the Gamma 1 ectosymbiont.

Figure 4.

RubisCO consensus tree based on ML analyses of deduced amino acid sequences of cbbL and cbbM genes (sequences from this study in red, in parentheses the number of sequences with > 95.0% to the given sequence). Scale bar = 0.10 estimated substitutions per site. Numbers next to nodes show to bootstrap values based on 100 ML replicates (only values above 60% are shown).

To examine the potential of the Epsilon 1 ectosymbiont for autotrophy, we examined the aclB gene, coding for the beta subunit of ATP citrate lyase. This gene is widespread in Epsilonproteobacteria including those commonly found at deep-sea hydrothermal vents but is not known from Gammaproteobacteria (Campbell et al., 2006; Nakagawa and Takai, 2008). We were not able to amplify this gene from DNA extracts of T. benedii tail ends, paralleling the lack of amplification products for the indicator gene soxB for sulfur oxidation in Epsilonproteobacteria. The metabolism of T. benedii's epsilonproteobacterial ectobionts therefore remains unclear. Epsilonproteobacteria are highly versatile and can live autotrophically, as well as mixo- or heterotrophically (Campbell et al., 2006). In the organic and sulfide-rich sediments in which T. benedii lives, any of these metabolisms are conceivable.

Nature of the T. benedii ectosymbiont association

The ectosymbiotic community associated with T. benedii has the potential for chemoautotrophic sulfur oxidation and it is possible that the symbionts fix inorganic carbon to organic carbon compounds that they pass on to their host. This has been suggested (but rarely proven) for other hosts with chemoautotrophic ectosymbionts such as stilbonematinid nematodes (Polz et al., 1992; Ott et al., 2004), the vent shrimp R. exoculata (Rieley et al., 1999; Zbinden and Cambon-Bonavita, 2003), the vent barnacle V. osheai (Suzuki et al., 2009), the vent crab S. crosnieri (Watsuji and Takai, 2009) and the cave amphipod N. ictus (Dattagupta et al., 2009). As mixo- or heterotrophs, the T. benedii ectosymbionts could use the waste products the host excretes during anaerobic metabolism such as succinate, acetate and propionate (Dubilier et al., 1994) and recycle these back to the host as suggested for gutless oligochaetes (Dubilier et al., 2001; Woyke et al., 2006). They could also take up organic compounds directly from the environment and pass these or essential amino acids and vitamins to the host. However, T. benedi has a fully functional digestive system and like other aquatic oligochaetes ingests sediment and the organic material and microorganisms therein to gain nutrition. It is therefore hard to imagine that its ectobiotic community can play an important role in its nutrition in comparison to the organic matter available in its surroundings. Furthermore, only the basal cells of the filamentous ectosymbionts are in direct contact with the worm, and thus provide very little surface area for the exchange of organic compounds.

Another function hypothesized to play a role in symbioses with sulfur-oxidizing bacteria is that these detoxify hydrogen sulfide, a potent inhibitor of aerobic respiration, for their hosts (Somero et al., 1989; Dattagupta et al., 2009). However, our calculations show that diffusion rates of sulfide through the body wall of the worms are so fast that bacterial sulfide oxidation would not be able to outcompete it (Dubilier et al., 1995a). Finally, it is also unlikely that the ectosymbionts are involved in pathogenic interactions, as we saw no differences in the vitality of worms lightly and heavily covered with the filaments in this study and an earlier one (Dubilier, 1986).

While the benefit of the ectosymbionts for T. benedii remains unclear, the advantages for these in associating with the worm are obvious. The sulfur-oxidizing ectosymbionts have access to an ideal environment on the worm's tail with access to sulfide rising up from the sediments and oxygen available from the water above the sediment surface. There is a strong selective advantage for sulfur-oxidizing bacteria in associating with animals that can provide them with both sulfide and oxygen. For example, Røy and colleagues (2009) showed that the sulfur-oxidizing symbionts of the ciliate Zoothamnium niveum take up 100 times more sulfide than bacteria on flat inert surfaces. This selective advantage is so strong that these types of associations have evolved multiple times in numerous lineages of sulfur-oxidizing bacteria and in a wide array of host groups (Dubilier et al., 2008). For the mixo- and heterotrophic members of the T. benedii microbial community, the association with an animal that excretes large amounts of carbon and nitrogen waste compounds provides a rich source of nutrition, and this advantage is hypothesized to play a role in many associations between bacteria and marine invertebrates (Carman and Dobbs, 1997; Robidart et al., 2008).

Specificity of the association

The T. benedii Gamma 1 and Epsilon 1 ectosymbionts belong to clades that consist almost exclusively of bacteria associated with hydrothermal vent invertebrates. This suggests that bacteria from these clades have developed an adaptive trait that enables them to easily colonize marine invertebrates. In the symbiotic associations between bioluminescent Vibrio fischeri and their marine hosts, Mandel and colleagues (2009) hypothesize that only a single regulatory gene was needed to confer free-living V. fisheri with the ability to colonize their hosts, possibly by ‘switching “on” pre-existing capabilities for interacting with an animal’. Intriguingly, the closest cultured relative to the T. benedii Gamma 1 ectosymbiont, the filamentous sulfur oxidizer L. mucor, forms a basal node to the γ-ectobiont clade (Fig. 1). Leucothrix mucor has been described from surfaces as diverse as marine algae, fish eggs, and dead and live aquatic invertebrates (Johnson et al., 1971; Sieburth, 1975; Carman and Dobbs, 1997; Payne et al., 2007), suggesting a less specific interaction between these bacteria and the surfaces they colonize. This could indicate that there is a progression from these possibly ancestral L. mucor associations to the highly stable and specific associations within the γ-ectobiont clade. However, given that most studies used only morphological characteristics to identify the filamentous epibionts as L. mucor, unambiguous identification through 16S rRNA sequencing and FISH is needed to test this hypothesis.


Symbiotic associations with Epsilonproteobacteria have only been described from deep-sea invertebrates, making T. benedii, to our knowledge, the first invertebrate from shallow environments with symbiotic Epsilonproteobacteria. Tubificoides benedii is an opportunistic species that is well adapted to estuaries and mudflats with organic-rich sediments and rapid environmental fluctuations, including low oxygen and high sulfide concentrations (Giere, 2006). Similarly, Epsilonproteobacteria have been described as uniquely suited to thrive in extreme environments such as deep-sea hydrothermal vents, where they can rapidly colonize dynamic environments with suboxic to anoxic conditions (Campbell et al., 2006). Thus, T. benedii and its epsilonproteobacterial ectosymbionts make good partners in sharing similar ecological niches.

Symbiotic associations with Gammaproteobacteria are widespread in both shallow-water and deep-sea chemosynthetic environments (Dubilier et al., 2008). However, the T. benedii Gamma 1 ectosymbiont belongs to a clade that consists exclusively of symbiotic and free-living bacteria from deep-sea hydrothermal vents, while the chemosynthetic symbionts of other hosts from shallow waters belong to clades distant from the γ-ectobiont clade (Fig. 1). Thus, T. benedii is, to date, unique in its symbiotic association with gamma- and epsilonproteboacterial symbionts related to those from deep-sea hydrothermal vents.

The discovery of the T. benedii ectosymbiosis shows that factors other than biogeography and host affiliation must have been the driving force behind the associations within the γ- and ε-ectobiont clades. Instead, the environment appears to have been crucial for the establishment and evolution of these ectosymbiotic associations, namely highly dynamic environments with strong fluctuations of oxidants and reductants. We hypothesize that symbioses with bacteria from these clades may be more widespread in shallow-water environments such as sulfide-rich intertidal mud flats than currently recognized. Morphological and molecular analyses of ectobionts from coastal marine sediments combined with analyses of their metabolic potential will be useful in providing a better understanding of the factors defining distribution patterns and function of associations between marine hosts and their symbiotic bacteria.

Experimental procedures

Specimen collection

Tubificoides benedii were collected in 1998 from Wadden Sea sediments at the Lister Haken in the Königshafen Bay on the Island of Sylt (55.03 N 8.10 E). The collection site is characterized by eutrophication and in the warmer months, massive green algal mats cover the sediment and lead to high sulfate reduction rates and sulfide concentrations (Kristensen et al., 2000; Reise and Kohlus, 2008).

To ensure as little contamination as possible, worms were allowed to defecate their gut contents before fixation, and only worms with visibly clear guts were used. Individual specimens were rinsed three times in 0.2 µm filtered seawater and fixed as described below.

Electron microscopy

For TEM specimens were fixed and prepared as described previously (Dubilier, 1986).

Cloning and sequencing of 16S rRNA, aprA, soxB, cbbL, cbbM and aclB genes

The tail ends of four T. benedii individuals were prepared individually for PCR as described previously (Dubilier et al., 1999) using the isolation protocol of Schizas and colleagues (1997). Briefly worms were digested with Proteinase K and DNA was extracted with Gene Releaser (BioVentures, Murfreesboro, TN, USA). Amplification, cloning and sequencing of the 16S rRNA, cbbL, cbbM and aprA genes was carried out as described previously (Blazejak et al., 2006) with the following modifications: only 28–30 cycles and an extra reamplification procedure of five cycles. The annealing temperature for cbbL was set to 56°C instead of 48°C, for aprA to 58°C instead of 54°C. For the amplification of cbbM, the primers cbbMF_Els (Elsaied and Naganuma, 2001) and a modified cbbM1R (Blazejak et al., 2006) with the sequence 5′ MGA GGT SAC SGC RCC RTG RCC RGC MCG RTG 3′ were used with an annealing temperature of 62°C. PCR of the soxB gene was carried out with soxB1446b and soxB432f with an annealing temperature of 47°C as described (Petri et al., 2001). For aclB the primer combinations used were aclB892F or aclB275F with aclB1204R with an annealing temperature of 42°C (Campbell et al., 2003; Takai et al., 2005).

For all genes, PCR products were purified with the QIAquick PCR purification kit (QIAGEN, Hilden, Germany) and cloned using the pGEM-T/pGEM-T Easy Kit (Promega, Madison, WI, USA) or the TOPO Kit (Invitrogen, Paisley, UK) according to the manufacturers' protocols. Plasmid DNA was purified from overnight cultures using the QIAprep plasmid kit (QIAGEN, Hilden, Germany). Clones with correct insert size were partially sequenced (approximately 450–900 bp) and grouped according to phylogentic positioning and similarity values from a distance matrix in ARB (Ludwig et al., 2004).

Sequencing reactions were run on the ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). For each host individual, at least one representative clone from each dominant clone group was fully sequenced in both directions. Full sequences within each clone group shared at least 99.0% sequence similarity (% identical nucleotides) for the 16S rRNA gene and at least 95% sequence similarity (% identical amino acids) for the protein coding genes. Closely related sequences of representative sequences were identified with BLAST (Altschul et al., 1990) queries and through phylogenetic analyses.

Phylogenetic analyses

16S rRNA chimeras were identified using CHIMERA_CHECK from the Ribosomal Database Project (Cole et al., 2007) and by eye in sequence alignments, and trimmed or excluded from further analyses. Alignments were based on the 16S rRNA secondary structure without partitioning into stem and loops (Pruesse et al., 2007). Phylogenetic analyses were performed with the ARB program (Ludwig et al., 2004) and the online version of RaxML (Stamatakis et al., 2008). For the 16S rRNA gene, phylogenetic trees were calculated with sequences no shorter than 1300 bp using neighbour-joining, parsimony and maximum likelihood (ML, HKY substitution model) as well as RaxML. Group filters for ML calculations (25%-, 30%-, 40%- and 50%-filters) were constructed from published gamma- and epsilonproteobacterial sequences in ARB. Bootstrap values for the gamma- and epsilonproteobacterial 16S rRNA trees were based on 1000 ML bootstraps calculated with ARB. Short sequences in the CFB and deltaproteobacterial 16S rRNA trees (Figs S1 and S2) were added to the ML-tree using parsimony with a positional variability filter, bootstraps resulted from whole tree ML calculations. Phylogenetic trees of protein-coding genes were generated from translated gene sequences of 131 (aprA), 133 (cbbM) and 230 (cbbL) amino acids using the ML algorithm and a JTT model with a 25% positional conservation filter. Trees were reconstructed using the standard operating procedure for phylogenetic inference SOPPI (Peplies et al., 2008), by visually comparing different methods, parameters and filters to identify the most stable tree topologies. All trees shown in this study were constructed based on ML analyses with nodes that were not stable (i.e. differed in more than two methods) collapsed to a consensus branch.

Accession numbers

The sequences from this study are available through GenBank under the accession numbers GU197394-GU197482.

Fluorescence in situ hybridization (FISH)

Worm tail ends were fixed, embedded and sectioned as described previously for gutless oligochaetes (Dubilier et al., 1995b) with the omission of the redundant postfixation step with 4% paraformaldehyde during the rehydration process. Sections of T. benedii posterior ends were prepared for FISH with monolabeled cy3 and cy5 probes and catalyzed reporter deposition (CARD) FISH with horseradish peroxidase labeled probes and tyramide signal amplification (with the fluorescent dyes Alexa 488 and 633) as described previously (Blazejak et al., 2006) with the following modifications: To decrease the loss of FISH signal in the mucous layer of the worm, additional digestion procedures were added. After 12 min of 0.3 M HCl, instead of a 5 min digestion with Proteinase K, the slides were immersed for 30 min to 1 h in 0.1% lysozyme in 0.1 M Tris/HCl / 0.05 M EDTA, 30 min to 1 h in 0.005% amylase in 1 × PBS (60 U ml−1) and 5 min in 0.0005% Proteinase K in 20 mM Tris/HCl (all enzyme incubations at 37°C). Washing for CARD-FISH in 1 × SSC buffer was increased to 1 h. Probe concentrations in the hybridization buffer were 3.3 ng µl−1 in FISH and 0.05 ng µl−1 in CARD-FISH)

Specific and group oligonucleotide probes targeting the dominant gamma- and epsilonproteobacterial 16S rRNA sequences found in T. benedii clone libraries (Table 1) were created with the ARB program and checked against sequences in GenBank with BLAST and in the RDP database with Probe Match (Cole et al., 2007) (Table 2). The specificity of the ectosymbiont probes was tested against reference bacteria with 16S rRNA sequences containing one or more mismatches unless otherwise noted. General probes for the Bacteria, Gammaproteobacteria, Deltaproteobacteria and the CFB served as positive controls and the nonsense probe NON338 as a negative control. Hybridizations were performed at formamide concentrations ensuring specificity for the targeted groups (Table 2). Images were recorded with the Axiovision camera (Zeiss) and optimized with the accompanying program AxioVision LE 4.5.


We thank Sabine Lenk for her help with soxB amplifications, and Silke Wetzel and Lisa Drews for excellent technical assistance. We are grateful to Claudia Bergin, Jan Küver and Sabine Lenk for fruitful discussions. This work was supported by the Max Planck Society, Munich, Germany.