Acquisition of a Novel Sulfur-Oxidizing Symbiont in the Gutless Marine Worm Inanidrilus exumae

ABSTRACT Gutless phallodrilines are marine annelid worms without a mouth or gut, which live in an obligate association with multiple bacterial endosymbionts that supply them with nutrition. In this study, we discovered an unusual symbiont community in the gutless phallodriline Inanidrilus exumae that differs markedly from the microbiomes of all 22 of the other host species examined. Comparative 16S rRNA gene sequence analysis and fluorescence in situ hybridization revealed that I. exumae harbors cooccurring gamma-, alpha-, and deltaproteobacterial symbionts, while all other known host species harbor gamma- and either alpha- or deltaproteobacterial symbionts. Surprisingly, the primary chemoautotrophic sulfur oxidizer “Candidatus Thiosymbion” that occurs in all other gutless phallodriline hosts does not appear to be present in I. exumae. Instead, I. exumae harbors a bacterial endosymbiont that resembles “Ca. Thiosymbion” morphologically and metabolically but originates from a novel lineage within the class Gammaproteobacteria. This endosymbiont, named Gamma 4 symbiont here, had a 16S rRNA gene sequence that differed by at least 7% from those of other free-living and symbiotic bacteria and by 10% from that of “Ca. Thiosymbion.” Sulfur globules in the Gamma 4 symbiont cells, as well as the presence of genes characteristic for autotrophy (cbbL) and sulfur oxidation (aprA), indicate that this symbiont is a chemoautotrophic sulfur oxidizer. Our results suggest that a novel lineage of free-living bacteria was able to establish a stable and specific association with I. exumae and appears to have displaced the “Ca. Thiosymbion” symbionts originally associated with these hosts. IMPORTANCE All 22 gutless marine phallodriline species examined to date live in a highly specific association with endosymbiotic, chemoautotrophic sulfur oxidizers called “Ca. Thiosymbion.” These symbionts evolved from a single common ancestor and represent the ancestral trait for this host group. They are transmitted vertically and assumed to be in transition to becoming obligate endosymbionts. It is therefore surprising that despite this ancient, evolutionary relationship between phallodriline hosts and “Ca. Thiosymbion,” these symbionts are apparently no longer present in Inanidrilus exumae. They appear to have been displaced by a novel lineage of sulfur-oxidizing bacteria only very distantly related to “Ca. Thiosymbion.” Thus, this study highlights the remarkable plasticity of both animals and bacteria in establishing beneficial associations: the phallodriline hosts were able to acquire and maintain symbionts from two very different lineages of bacteria, while sulfur-oxidizing bacteria from two very distantly related lineages were able to independently establish symbiotic relationships with phallodriline hosts.


RESULTS AND DISCUSSION
Our morphological and molecular analyses revealed an unusual symbiotic community in Inanidrilus exumae, consisting of cooccurring gamma-, alpha-, and deltaproteobacterial symbionts ( Fig. 1 and 2; see Fig. S1 in the supplemental material). In the five host species whose symbiont communities have been examined so far, alpha-and deltaproteobacterial symbionts appeared to be mutually exclusive (15,16,21,22). Furthermore, we found no evidence for the presence of the primary symbiont, "Ca. Thiosymbion," in I. exumae. This is surprising because the 22 gutless phallodriline species examined to date have always harbored "Ca. Thiosymbion" symbionts (6). In contrast, I. exumae harbors a sulfur oxidizer that resembles "Ca. Thiosymbion" in appearance and function but belongs to a lineage of Gammaproteobacteria not previously known to be associated with gutless phallodrilines or other eukaryotic hosts. In the following discussion, we will focus on the morphology, phylogeny, and potential function of this novel gammaproteobacterial symbiont of I. exumae. A brief description of the phylogenies and possible functions of the secondary alpha-and deltaproteobacterial symbionts of I. exumae is provided in the supplemental material.
Morphology and phylogeny of the Gamma 4 symbiont. Only a single gammaproteobacterial 16S rRNA phylotype, which we named Gamma 4, was found in the clone libraries from I. exumae worms ( Table 1). Out of a total of 734 sequenced clones from seven host individuals, we never found a sequence that belonged to the "Ca. Thiosymbion" clade. In previous studies of other gutless oligochaete species, "Ca. Thiosymbion" could always be amplified with the general 16S rRNA primers we used in this study (8F and 1492R) (21,22).
Fluorescence in situ hybridization (FISH) provided further support for our assumption that the Gamma 4 symbiont is the only gammaproteobacterium present in I. exumae. FISH with a probe specific to the 16S rRNA gene sequence of the Gamma 4 symbiont ( Table 2, IexuGAM4) showed that this sequence originated from large, oval-shaped bacteria (2 to 3 m long and 1 to 2 m wide) that were highly abundant and dominated the symbiont-containing region in all host individuals examined ( Fig.  1A and D). Dual FISH hybridization with the specific IexuGAM4 probe and the general probe for Gammaproteobacteria ( Table 2, GAM42a) showed a complete overlay of the hybridization signals, with both probes hybridizing in cells of the same large, ovalshaped morphotype (Fig. 1D). These results indicate that the Gamma 4 symbionts are the only Gammaproteobacteria present in I. exumae and that these hosts lack the "Ca. Thiosymbion" symbionts found in all other gutless phallodriline species examined.
Transmission electron microscopy (TEM) showed that the ultrastructure of the Gamma 4 symbionts was remarkably similar to that of "Ca. Thiosymbion" symbionts ( Fig. 1I and J). Like "Ca. Thiosymbion," the Gamma 4 symbiont was the largest (2-to 3-m) and most abundant morphotype of the symbiotic community, and its cells were also filled with large, electron-dense globules (Fig. 1I).
Comparative phylogenetic analyses of 16S rRNA gene sequences revealed that the Gamma 4 symbiont belongs to a novel lineage of Gammaproteobacteria not previously shown to be associated with animal or plant hosts (Fig. 2). While the phylogenetic resolution of the gammaproteobacterial tree was not well defined at the basal nodes ( Fig. 2), we never observed any clustering of the Gamma 4 symbiont sequence with the "Ca. Thiosymbion" clade in our analyses. Indeed, the 16S rRNA gene sequence of the Gamma 4 symbiont differed from sequences belonging to the "Ca. Thiosymbion" clade by more than 10% (Fig. 2). The closest uncultured relative, with a sequence divergence of 7%, was a sediment clone from a beach in the Cíes Islands off the coast of northern Spain (GenBank accession number JF344692). The closest cultured relatives, with sequence divergences ranging from 9 to 10%, were sulfur-storing members of the family Ectothiorhodospiraceae and bacteria from the genera Nitrosococcus and Methylococcus.
Indications for autotrophic sulfur oxidation by the Gamma 4 symbiont. Despite their divergent phylogenies, "Ca. Thiosymbion" and the I. exumae Gamma 4 symbionts not only share highly similar morphologies but also appear to have similar functional roles as chemoautotrophic sulfur oxidizers. As shown for other chemoautotrophic symbioses, the cbbL gene, coding for one of the key proteins of the Calvin-Benson-Bassham (CBB) cycle, the ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO) form I large subunit, was present in I. exumae ( Fig. 3A and Table 1) (23). The cbbL sequence obtained from I. exumae grouped with sequences from other gammaproteobacterial chemoautotrophs, such as free-living Chromatiaceae and sulfur-oxidizing symbionts from other marine invertebrates. It is therefore likely that the I. exumae cbbL sequence originated from the I. exumae Gamma 4 symbiont.
Evidence for the potential of the Gamma 4 symbiont to oxidize reduced sulfur compounds was provided by Raman spectroscopy analyses, which revealed sulfur in the cells of these symbionts ( Fig. 1K; Fig. S2.1 and S2.2). Moreover, we amplified aprA genes (encoding AprA, the alpha subunit of adenosine-5=-phosphosulfate [APS] reductase) related to those of free-living and symbiotic sulfur-oxidizing bacteria from I. exumae individuals ( Fig. 3B and Table 1). Sequences belonging to two phylogenetically distinct APS reductase lineages, AprA I and II, were found in I. exumae (Fig. 3B). We assume that the sequences from both AprA I and II originated from the Gamma 4 symbiont, as no other gammaproteobacterial sulfur oxidizers were found in I. exumae and the alphaproteobacterial symbionts of gutless phallodrilines do not appear to have an APS reductase (22). The presence of two gene loci for AprA has been shown for several free-living sulfur-oxidizing bacteria and is therefore not unusual (24). Meyer and Kuever (24) hypothesized that the presence of two gene loci might provide physiological versatility in habitats with oscillating oxygen and sulfide concentrations. This may well be the case for I. exumae and other gutless phallodrilines, which migrate between upper, oxidized and lower, sulfidic sediment layers.
Symbiont replacement in I. exumae? What are the evolutionary events that might explain the presence of a novel sulfur-oxidizing symbiont and the absence of the ubiquitous "Ca. Thiosymbion" in I. exumae? "Ca. Thiosymbion" is present in all 22 gutless phallodriline species examined to date from habitats around the world, including six host species from the Bahamas, some of which cooccur with I. exumae (6,20). All "Ca. Thiosymbion" 16S rRNA gene sequences are closely related to each other and belong to a monophyletic clade (6). The phallodriline hosts have also evolved from a single common ancestor, based on morphological (25,26) and molecular data (27,28). Furthermore, I. exumae is not an early-diverging or basal species within the gutless phallodrilines but, rather, closely related to other Inanidrilus species, which form a monophyletic group within the gutless phallodrilines ( Fig. 4). Since all gutless phallodrilines, including the four Inanidrilus species closely related to I. exumae (Fig. 4), harbor "Ca. Thiosymbion" symbionts (6,14), the most parsimonious conclusion is that the ancestor of I. exumae also harbored a "Ca. Thiosymbion" symbiont.
How could the Gamma 4 symbiont have displaced "Ca. Thiosymbion" in I. exumae? We envision the following three successive scenarios that could explain how the ancestral symbiont of I. exumae was displaced. In the first step, when "Ca. Thiosymbion" was still the primary symbiont, the ancestors of the Gamma 4 symbiont must have been able to enter and persist in I. exumae at some point in their evolutionary history. The early developmental stages of the worms were the most likely window of opportunity for infection by bacteria from the environment. Gutless phallodrilines lay single eggs into the surrounding sediment. The egg remains attached to the parent worm and is fertilized with sperm and coated with symbiotic bacteria from the parent worm in a smear infection and then encased in a cocoon, which is eventually deposited in the sediment (17,18). Free-living bacteria from the sediment could easily become encased within the cocoon during this process and colonize the developing embryo.
During a second, transition phase, the Gamma 4 bacteria and the "Ca. Thiosymbion" symbiont may have coexisted in I. exumae. In some gutless phallodriline species, "Ca. Thiosymbion" cooccurs with secondary sulfur-oxidizing Gammaproteobacteria, called Gamma 2 and 3 symbionts (15,16). However, these secondary sulfur-oxidizing symbi- onts are much smaller than "Ca. Thiosymbion" and occur in the small interstitial spaces between the large "Ca. Thiosymbion" cells, so that competition for space does not appear to occur. Also, they have functional differences that may allow niche separation: the Gamma 3 symbiont of Olavius algarvensis, for example, uses nitrate as an electron acceptor, while "Ca. Thiosymbion" uses oxygen (12,13). Furthermore, the O. algarvensis Gamma 3 symbiont can use additional electron donors, such as carbon monoxide, which cannot be used by its "Ca. Thiosymbion" symbiont, thereby reducing competition for energy sources (13,29).
In the third and final step of displacement, the Gamma 4 symbiont appears to have outcompeted "Ca. Thiosymbion" in I. exumae, at least in the host population we examined (16 individuals from the same collection site were examined with molecular methods or FISH). While it is possible that our methods were not sensitive enough to a Sequences that shared Ͼ99% identity were grouped as a single phylotype. One or more clones of each phylotype and individual were sequenced in both directions for the almost-full-length 16S rRNA gene sequence and for partial aprA and cbbL gene sequences. SRB, sulfate-reducing bacteria. b -, not analyzed. detect residual, very low numbers of "Ca. Thiosymbion" cells in the I. exumae individuals we examined, these hosts were clearly dominated by Gamma 4 symbionts. The niche separation between "Ca. Thiosymbion" and Gamma 4 symbionts may not have been sufficient to allow their codominance in I. exumae. However, other factors could also explain the displacement of "Ca. Thiosymbion," such as a massive viral infection event, a strong competitive advantage of the Gamma 4 symbiont over "Ca. Thiosymbion," or harmful mutations in the ancestral "Ca. Thiosymbion" population. Recent studies have shown that symbiont displacement is not as rare as previously assumed. Even in associations in which vertical transmission of symbionts occurs over long evolutionary times, acquisition of symbionts from novel lineages of environmental bacteria and symbiont displacement can occur occasionally in both aquatic and terrestrial symbioses (30)(31)(32)(33)(34)(35)(36)(37)(38). In the gutless phallodriline symbioses, Zimmermann et al. (6) revealed that displacement of "Ca. Thiosymbion" may have occurred numerous times. However, in the 22 phallodriline species analyzed by Zimmermann et al. (6), displacement appears to have always occurred within the "Ca. Thiosymbion" clade; that is, the ancestral "Ca. Thiosymbion" strain of a given host species was displaced by a "Ca. Thiosymbion" strain from another host species. I. exumae is the only species in which Based on close phylogenetic relationships to free-living and symbiotic sulfur-oxidizing bacteria, we assume that the cbbL sequence and the aprA sequences from AprA lineages I and II originated from the Gamma 4 symbiont, while a fifth aprA sequence most likely originated from a deltaproteobacterial symbiont. Asterisks show bacteria that have aprA gene sequences from both lineage I and II. Sequences obtained in this study are framed with a red box, sequences from "Ca. Thiosymbion" are highlighted in purple, and sequences from other gutless phallodriline symbionts are highlighted in yellow. GenBank accession numbers are shown. Scale bars represent 10% estimated phylogenetic divergence for nonmultifurcated branches. Numbers in the polygons show the number of bacterial species concatenated in the node. Black-or white-filled circles indicate maximum-likelihood bootstrap values as indicated in the keys, while percentages show posterior probabilities from Bayesian inference.
we found indications for the displacement of "Ca. Thiosymbion" by a novel, phylogenetically distinct lineage of bacteria not closely related to "Ca. Thiosymbion." Genomic, transcriptomic, and proteomic analyses of the Gamma 4 symbionts are needed to better understand the factors that allowed these bacteria to successfully colonize and persist in I. exumae.

MATERIALS AND METHODS
Site description and specimen collection. Inanidrilus exumae specimens were collected from shallow water sediments off Lee Stocking Island, Bahamas, in April 1999. I. exumae cooccurred with several other gutless phallodriline species in a water depth of about 3 m in sediments that were largely composed of fine calcareous sands (20). The worms were extracted by decantation and identified under a microscope. In total, 16 specimens were divided for different analyses: 8 were fixed in 80% ethanol for DNA extraction (7 for analysis of bacterial genes and 1 for analysis of host genes), and another 8 were cut and fixed either for TEM or for FISH as described previously (16,39). Samples were stored at 4°C. DNA preparation and PCR amplification. For DNA extraction and subsequent PCR of bacterial genes, seven individual worms were prepared singly. Specimens were rinsed in MilliQ water, and DNA was isolated as described previously (6,21), following the method of Schizas and colleagues (40). The bacterial 16S rRNA genes were amplified with primers specific for the bacterial 16S rRNA gene 8F and 1492R (41) using Taq DNA polymerase (Eppendorf, Hamburg, Germany). The bacterial 16S rRNA genes from I. exumae individuals 1 and 2 were amplified by applying the reconditioning approach (42,43) under the following conditions: initial denaturation at 96°C for 5 min, 15 plus 5 and 15 plus 7 cycles for I. exumae 1 and I. exumae 2, respectively, at 96°C for 1 min, 44°C for 2 min, and 72°C for 3 min, followed by a final elongation of 10 min at 72°C. The PCR conditions for I. exumae individuals 3, 4, and 5 were as described previously (22). The PCR conditions for I. exumae individuals 6 and 7 were initial denaturation at 94°C for 5 min, 30 cycles at 94°C for 1 min, 42°C for 1.5 min, and 72°C for 2 min, followed by a final elongation of 30 min at 72°C. The PCR protocols differed due to protocol improvements in the course of our biodiversity studies during the last decade and sample availability.
Genes coding for RubisCO form I and APS reductase were PCR amplified with 30 and 33 cycles, respectively. The following primers were used: cbbLF (5=-CACCTGGACCACVGTBTGG-3=) and cbbLR

FIG 4
Phylogenetic tree of gutless phallodrilines using Bayesian inference analysis of six concatenated genetic markers for the host (mt12S, mt16S, 18S, 28S rRNA, mtCOI, and ITS genes). Posterior probability values are indicated at nodes. Scale bar represents 10% estimated phylogenetic divergence for nonmultifurcated branches. GenBank accession numbers for I. exumae sequences are given in the figure and under "Accession number(s)" in the text. GenBank accession numbers for species with asterisks are as follows: 18S rRNA, KP943792 to KP943817; 28S rRNA, KP943818 to KP943844; mtCOI, KP943845 to KP943866; ITS, KP943867 to KP943884; mt12S rRNA, KP943885 to KP943908; and mt16S rRNA, KP943909 to KP943931.
Host genes were amplified and sequenced from DNA extracted from a single I. exumae individual (sample CE73) as previously described (6).
Cloning and sequencing. PCR products for all bacterial genes (16S rRNA, cbbL, and aprA) were cloned separately for each individual worm using the pCR4-TOPO plasmids and TOP10 chemically competent cells (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Clones were selected for the correct insert size and sequenced, and sequences grouped in clone groups as described in reference 44. PCR products for amplified host genes were sequenced directly.
Phylogenetic analyses of symbiont sequences. Sequences were checked with BLAST (45, 46) for similarity searches. Chimeras were identified using CHIMERA_CHECK from the Ribosomal Database Project (RDP) (47) and manually in sequence alignments and were excluded from further analysis.
Sequences were trimmed at the 5= and 3= ends, and only nearly full-length 16S rRNA gene sequences, including outgroup sequences, were considered for tree calculations (Ͼ1,200 bp) using the ARB software package (48) and SILVA SSU Ref, release_NR99_119 July 2014 (49). The sequence similarities of the nucleotide sequences were calculated by distance matrix analysis, excluding the primer region. Phylogenetic trees for 16S rRNA gene sequences were calculated using Bayesian inference (MrBayes version 3.2) (50) and maximum-likelihood (ML)-based methods (PHYML) provided within the ARB software package as described previously (6). We used the generalized time reversible (GTR) substitution model for both analyses. Trees for alpha-, gamma-, and deltaproteobacterial symbionts were calculated separately, and consensus trees were constructed based on the information from the Bayesian inference and maximum-likelihood analyses. Node stability was evaluated using posterior probabilities (Bayesian inference).
The phylogenies of the aprA and cbbL genes were generated from partial sequences of deduced amino acid sequences, with 134 and 101 amino acid positions compared, respectively. Sequences for each gene were aligned separately using MAFFT, provided within the ARB software package, and the 5= and 3= ends trimmed. For phylogenetic tree reconstruction, we used maximum-likelihood analyses (PHYML with LG and RAxML with JTT) and the bootstrapping algorithm in RaxML (51), as well as Bayesian inference (50). For the Bayesian inference analyses, the optimal model of amino acid evolution for AprA and CbbL was determined using ProtTest3 (https://github.com/ddarriba/prottest3) (LGϩG for both proteins). The protein alignments were imported into MrBayes version 3.2 and run in duplicate runs with four chains each (one hot and three cold) until convergence (26 million generations for AprA and 50 million generations for CbbL). Trees were sampled every 1,000 generations and were then summarized in a majority rule consensus using a burn-in value of 20%. Clade posterior probabilities were plotted onto the ML trees shown in Fig. 3.
Phallodriline host phylogeny. The mitochondrial 12S (mt12S), mt16S, and mtCOI, nuclear 18S and 28S rRNA, and ITS genes of 22 gutless phallodrilines and 5 gut-bearing annelids submitted by Zimmermann et al. (6) and the genes from I. exumae [see "Accession number(s)" below] were used for phylogenetic reconstruction. Sequences for each gene were aligned separately using MAFFT version 7 (52) with the Q-INS-I setting (53), alignments were manually adjusted, and the 5= and 3= ends trimmed using BioEdit as described in Zimmermann et al. (6). The optimal substitution model for each alignment was assessed, and phylogenetic trees were reconstructed using Bayesian inference (MrBayes version 3.2) (50) as described previously (6). Node stability was evaluated using posterior probabilities (Bayesian inference) and bootstrap support (100 RaxML rapid bootstrap runs), with values above 0.80 considered significant.
FISH. Parts of eight I. exumae individuals were fixed and prepared for fluorescence in situ hybridization (FISH) as described previously (16), with the slight modification that we used xylol instead of Roti-Histol (Carl Roth, Karlsruhe, Germany). Symbionts were detected by catalyzed reporter deposition (CARD)-FISH as described previously (54), with slight modifications as follows. Tissue sections were hybridized with the horseradish peroxidase (HRP)-labeled probe for 2.5 h at 46°C. After washing for 15 min at 48°C in washing buffer, the sections were equilibrated for 20 min at room temperature in phosphate-buffered saline (PBS; pH 8.0). The moist tissue sections were incubated with amplification solution (1ϫ PBS, pH 8.0, 2 M NaCl, 0.1% blocking reagent in 100 mM maleic acid buffer, pH 7.5, 0.0015% [vol/vol] H 2 O 2 , and 1% Alexa Fluor 488, 546, or 633 dye [Molecular Probes, Leiden, The Netherlands]) for 30 min at 46°C in the dark and rinsed in 1ϫ PBS buffer for at least 20 min at room temperature. For dual and triple hybridizations, the CARD-FISH protocol was repeated two or three times on the same sections using different probes and Alexa Fluor dyes, and the HRP was inactivated after each hybridization round by using 0.01 M HCI for 10 min at room temperature after the last washing step (16).
The oligonucleotide probes and formamide concentrations used in this study are listed in Table 2. Probes designed with ARB were checked for in silico specificity against sequences in GenBank using BLAST and against rRNA sequence databases using ProbeCheck (55). The specificity was also tested experimentally against mismatched 16S rRNA gene sequences of either reference strains or symbionts. General probes for Bacteria (EUB338 I to III), Gammaproteobacteria (GAM42a), and a subgroup of the Deltaproteobacteria (DSS658) were used as positive controls, and the antisense probe NON338 was used as a negative control. All hybridizations were performed at formamide concentrations ensuring the highest possible specificity.
TEM. Parts of eight I. exumae worms were fixed for transmission electron microscopy (TEM), washed in 0.05 M NA-cacodylate, and postfixed in osmium tetroxide. After dehydration in an acetone series, specimens were embedded in Spurr resin (56), and the worms' middle parts, containing the symbiont region, sectioned on an ultramicrotome. For electron microscopy, ultrathin sections were stained with uranyl acetate and lead citrate and examined with a Zeiss EM 902A (39).
Raman spectroscopy. Raman spectroscopy was done on parts of two of the individuals used for FISH analyses as described in Eichinger et al. (57). More details on material and methods, as well as results and discussion, can be found in the supplemental material.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AEM .02267-17.