Sampling: The “Puy-de-la-Poix” is a naturally flowing tar source located near the airport of Clermont-Ferrand (France, 45.7822 N 3.14642 E, 340 m, Fig. 1). The source appears as a small pool of ca. 3 m by 2 m, which may be covered by rain fall water according to the season, and overflows into the adjacent agricultural fields. At the time of sampling, the air temperature was 6 °C. The tar formed a thick mass on top of the source and the water covering the tar was ca. 2–5 cm deep. The water above the tar was sampled aseptically using sterile syringes (100 ml, surface water samples, SW). The remaining surface water was removed to avoid cross contamination with waters below the tar. Underground waters (UW, 100 ml) were sampled below the tar using sterile tubing punched through the tar. The chemical composition of the underground water can be found in Table S1. A 3-cm-thick, 25 × 25 cm square piece of bitumen was cut aseptically with a sterile blade (bitumen sample, B).

16S library construction: Metagenomic DNA was prepared from 1 g of bitumen or 50 ml of water sample as described previously [10]. The complete 16S rRNA gene was amplified using the universal eubacterial primers pA and pH [11] and cloned into pGEM-T easy (Stratagene, France) to generate the UW, B, and SW libraries. The complete sequence of the 16S gene was determined by Sanger sequencing. GenBank accession numbers for representative groups of Epsilonproteobacteria are JX403021-JX403066.

Diversity analyses: Sequences were classified using the RDP classifier with a confidence threshold of 80% (https://www.rdp.cme.msu.edu). Mothur 1.36 [12] was used to eliminate chimera (chimera.uchime), assign sequences to OTUs (align.seqs, dist.seqs, cluster, OTUs, 97%), get representative sequences (get.oturep), and calculate diversity values (H_Shannon and S_Chao1). Phylogenetic trees were reconstructed with Seaview using PhyML, and supported by 1000 boostrap repetitions [13].

Cultivation: Growth was assayed in media specific for Epsilonproteobacteria: Thiomicrospira medium DSM142, Sulfurospirillum medium DSM771 or Sulfurimonas medium DSM1053 (http://www.dmsz.de). Growth media were supplemented with electron donors (lactate, acetate, fumarate, formate, CO₂) and acceptors (SO₂, SO₄, NO₃, SeO₃, thiosulfate) at 5 mM final concentration. Twenty-milliliter enrichment cultures were set in sealed 50-ml serum vials under nitrogen atmosphere. Enrichment cultures were started with a 200-mg piece of bitumen. Cell counts were determined at least twice a week using a Thoma cell. Positive cultures were subcultured by serial dilution every two weeks until pure cultivable isolates were obtained. Purity was estimated by visual inspection of the culture and PCR amplification, cloning and sequencing of the 16S rRNA gene.

Bacterial diversity at the Puy-de-la-Poix. A total of 159 16S rDNA bacterial sequences, e.g., 58, 49 and 52 clones from the bitumen (B), underground water (UW) and surface water (SW) libraries, respectively, were obtained after quality and chimera screening. The sequences clustered into 48 different OTUs, belonging to six phyla (Table S2). The diversity is dominated by proteobacteria, e.g., 96,5%, 93,9%, and 84,6% in the B, UW, and SW libraries, respectively. The community diversity and richness is lowest in the bitumen (Table 1). Community richness is similar in both water bodies, although the H_shannon index evidences a higher diversity in the surface water. The higher diversity is associated with drastically different bacterial communities. Sequences of the SW library are composed of equal fractions of alpha-, beta-, epsilon and gammaproteobacteria (Fig. 2), while in the UW and B libraries are dominated by Epsilonproteobacteria (80.4 and 83.9% respectively). Most Epsilonproteobacteria sequences at Puy-de-la-Poix are associated with the genus Sulfurimonas, which represent 56.9% and 63.3% of the total bacterial diversity in the B and UW libraries, respectively. The remaining diversity of Epsilonproteobacteria is represented by Sulfurovum and Arcobacter (12.1% and 10.3% of the total bacterial diversity respectively in the B library, and 8.1% of the total bacterial diversity for both groups in the UW libraries). A single Sulfurimonas OTU, e.g. OTU 11 (Table 2, Fig. 2) dominates Epsilonproteobacteria diversity and accounts for 30 and 35% of Epsilonproteobacteria in the UW and B libraries (Fig. 2), but is not found in the SW library. OTU 11 is most closely related to the type species of the Sulfurimonas gotlandica species [14]. Four OTUs were associated with the Sulfurovum aggregans type strain Monchim33, which has been isolated from a hydrothermal vent system [15]. The Sulfurovum OTUs are related to environmental clones of very diverse origins, a soil, a sediment mesocosm, a bivalve symbiont or an acid cave. Arcobacter sequences (6 OTUs) represent a minor, but significant, component of the bacterial diversity (10% and 8% for the B and UW libraries respectively). The last two OTUs belong to the Sulfuricurvum and Sulfurospirillum genera (Table 2). There is a large overlap between the bacterial populations of the UW and B libraries, which is consistent with a common origin of the two populations, and the probability that Epsilonproteobacteria were present in the zone of contact of the oil and water. Indeed, such large prevalence of Epsilonproteobacteria has been reported previously at the oil/water interface in other geological settings such as a highly degraded oil reservoir or an oil spill from a storage tank [5,6]. The absence of a bacterial diversity related to known bacterial fermenters or syntrophs at the Puy-de-la-Poix, and the low level of sulfate-reducers suggested that the degradation of the organic matter of the oil and sulfate reduction may not be occurring at a significant level at the Puy-de-la-Poix, and that microbial degradation occurred during the transit of the oil to the surface.

Isolation of cultivableEpsilonproteobacteria. Obtaining cultivable Epsilonproteobacteria from the Puy-de-la-Poix was a prerequisite to assay their metabolic abilities. Six pure cultures were obtained using the Thiomicrospira mineral base medium supplemented with either acetate or lactate, and nitrate as an electron acceptor. The 16S rDNA gene sequences of the six isolates were almost identical (> 99% sequence identity), defining a single OTU closely related to Sulfurospirillum multivorans (Clone L12, Table 2 and Fig. 3). The strains isolated here are closely related, but do not belong to OTU 23 (SW library, Table 2, Fig. 3). Microscopic observations of the isolates revealed curved or helical rods. The cells were non-motile. Growth was obtained with acetate and lactate, but not pyruvate, formate or sugars in the presence of nitrate under anaerobic conditions. No growth was observed on sulfate or thiosulfate as electron acceptors. As expected, no growth occurred under aerobic conditions. Typical growth was low, reaching ca. 10⁷ cells/ml within two weeks. The apparent doubling time of strain L12 is ca. 24 h. This doubling time is extremely long for the Sulfurospirillum species, for which the doubling time is usually in the order of a few hours. Growth yields were also extremely low. Both observations indicate that the growth medium used to isolate strain L12, although supporting growth, is lacking some important components, which we have been unable to identify thus far. No pure culture belonging to the major Epsilonproteobacteria clades (Fig. 3) could be isolated.

Role ofEpsilonproteobacteriaat Puy-de-la-Poix. Only two OTUs observed at the Puy-de-la-Poix are related to isolates from oil/water interfaces: OTU 10 is related to Sulfuricurvum kujiense, a species isolated from an oil storage tank [16]. The second OTU is related to the CVO strain, which has been isolated from the Coleville oil field [17]. Both S. kujiense and CVO are able to reduce nitrate with the concomitant oxidation of sulfur compounds. Both organisms have been found to dominate the bacterial diversity at their isolation sites, although they were unable to degrade the organic matter of the oil. Therefore, their role in these particular ecosystems, and the source of carbon supporting their growth remained unclear. Most of the Epsilonproteobacteria at Puy-de-la-Poix were most closely related to Sulfurimonas gotlandica [14]. S. gotlandica is a sulfur oxidizer, nitrate-reducing chemolithoautotrophic bacterium that has been isolated from the deep-sea abyssal sediments of the Gotland depth. Similarly to S. kujiense and strain CVO, S. gotlandica is unable to use small organics, such as pyruvate, lactate of formate, but uses dissolved carbonates to support its growth, a phenomenon known as high dark CO₂ fixation [18]. Thus, it is very unlikely that Epsilonproteobacteria at Puy-de-la-Poix may be involved in the breakdown of the complex organic matter of the tar. High dark CO₂ fixation is important in the suboxic to sulfidic transition zones of aquatic pelagic redox zones, as in the Black Sea and the Baltic Sea from which S. gotlandica has been isolated [14,18–21]. There are enough carbonates in the water at Puy-de-la-Poix to support high dark CO₂ fixation by chemolithoautotrophic Epsilonproteobacteria (Table S2) oxidizing sulfur and reducing nitrates. Thus, the dominance of Epsilonproteobacteria could be explained by the refractory nature of the residual organic matter composing the tar, which would limit the access of fermenters and syntrophs to carbon and energy sources, while Epsilonproteobacteria could rely on reduced sulfur compounds and CO₂ fixation for growth. Epsilonproteobacteria could reduce sulfur compounds of the oil, but more probably dissolved reduced sulfur species, such as hydrogen sulfide, which are present in the water due to the hydrothermal activity of this geologic province.

The bacterial diversity at Puy-de-la-Poix shows an unexpected domination of Epsilonproteobacteria. The lack of known sulfate-reducers or syntrophs is a clear indication that oil mineralization to methane does not occur at this site. The abundance of Epsilonproteobacteria indicates that reduced sulfur species are the major energy source, ruling out significant degradation of the oil organic matter under anaerobic conditions. Our observations do not rule out the possibility of aerobic microbial oil degradation, in contact with surface waters, or when oil overflows in the nearby fields. The presence of large populations of Epsilonproteobacteria at the water/oil interface of different geological settings (oil reservoirs, oil contaminated waters or natural seep) is a clear indication that this interface constitutes a favorable ecological niche for these organisms.