Transformation of the recalcitrant pesticide chlordecone by Desulfovibrio sp.86 with a switch from ring-opening dechlorination to reductive sulfidation activity

The insecticide chlordecone has been used in the French West Indies for decades, resulting in long term pollution, human health problems and social crisis. In addition to bacterial consortia and Citrobacter sp.86 previously described to transform chlordecone into three families of transformation products (A: hydrochlordecones, B: polychloroindenes and C: polychloroindenecarboxylic acids), another bacterium Desulfovibrio sp.86, showing the same abilities has been isolated and its genome was sequenced. Ring-opening dechlorination, leading to A, B and C families, was observed as previously described. Changing operating conditions in the presence of chlordecone gave rise to the formation of an unknown sulfur-containing transformation product instead of the aforementioned ones. Its structural elucidation enabled to conclude to a thiol derivative, which corresponds to an undocumented bacterial reductive sulfidation. Microbial experiments pointed out that the chlordecone thiol derivative was observed in anaerobiosis, and required the presence of an electron acceptor containing sulfur or hydrogen sulfide, in a confined atmosphere. It seems that this new reaction is also active on hydrochlordecones, as the 10-monohydrochlordecone A1 was transformed the same way. Moreover, the chlordecone thiol derivative called F1 was detected in several chlordecone contaminated mangrove bed sediments from Martinique Island, highlighting the environmental relevance of these results.

www.nature.com/scientificreports/ In this context, chlordecone microbial degradation has been the focus of several studies: some Pseudomonas aeruginosa sp. as well as mixed cultures isolated from the Hopewell area in the 1980s exhibited significant aerobic dechlorination ability leading to mono-and dihydrochlordecones transformation products (TPs) 3 . Later on, the conversion of chlordecone into unknown polar and nonpolar TPs in the presence of the anaerobic archaea Methanosarcina thermophila at 50 °C was reported 4 . More recently, two bacterial consortia (86 and 92), isolated from microbial enrichment cultures from chlordecone-contaminated soils and wastewater treatment plant sludge, were able to transform chlordecone, under anoxic conditions, at room temperature. Several TPs were identified and grouped into three families: hydrochlordecones (family A), polychloroindenes (family B) and polychloroindenecarboxylic acids (family C) 5,6 . TPs from families B and C that were predominant arose from a ring-opening dechlorination of the chlordecone bishomocubane cage. Metagenomic analysis of these consortia indicated the prevalence of Citrobacter and Desulfovibrio sp. genomes among the dozen of bacterial genomes sequenced. Two members of these consortia, Citrobacter sp.86 and Citrobacter sp.92 that were isolated, both degraded chlordecone resulting in the same TP profile. Reductive dehalogenases are key enzymes in halorespiration 7,8 and could be indicative of a dehalogenation process. Here no gene sequences related to these enzymes were detected neither in the metagenomic data nor in the isolated Citrobacter genomes 5 .
The mechanistic degradation pathways and the exact role of each bacterium remains to be elucidated. In the past, several Desulfovibrio sp. or other sulfate-reducing bacteria have been reported in microbial cultures degrading chlorinated compounds [9][10][11][12][13][14] .
Herein, we report the isolation of Desulfovibrio sp.86 previously identified in chlordecone-degrading consortium 86 5 and explore its capacity to transform chlordecone and other chlorinated derivatives. TPs from the B and C families as well TP A1 were detected in pure cultures of Desulfovibrio sp.86. Unexpectedly, other incubation conditions in presence of chlordecone gave rise to the formation an unknown sulfur-containing TP instead of the aforementioned TPs. The elucidation of its structure allowed us to conclude to a thiol derivative. To the best of our knowledge, it corresponds to an unprecedented bacterial reductive sulfidation. This thiol derivative was detected in several chlordecone-contaminated mangrove bed sediments from the Martinique Island.

Results
isolation of Desulfovibrio sp.86. Isolation of Desulfovibrio sp.86 was achieved from the chlordeconedegrading consortium 86 5 using sulfate-reducing conditions. As bacterial consortium 86, able to transform chlordecone, was obtained from an enriched mineral medium (MM) named MM + 5 , supplemented with chlordecone, the main chemical composition was kept but electron donors and acceptors were modified. Several media formulations used to enrich sulfate-reducing bacteria utilize organic acids e.g. lactate, as carbon and energy sources (electron donor) and sulfate that is used as eletron-acceptor for growth [15][16][17][18] . In this context, pyruvate in the MM + liquid medium was replaced by lactate and sulfate was added (MMD liquid medium, see "Methods" section). The enrichment was spread on MMD agar and the brown vibrio-like bacterial colonies (observed under optical microscope) were purified further through two additional plate streakings (Fig. S1). An isolated bacterial strain was found to be identical to Desulfovibrio sp.86 from consortium 86 based on 100% identities of their 16S rRNA genes (1538 bp each).
Genome analysis of Desulfovibrio sp.86. The whole genome consists of a single 3,464,070 bp circular chromosome. CheckM analysis 19 performed with 61 genomes and 284 lineages indicates that the genome belongs to the Deltaproteobacteria and CheckM Completeness is 100% (zero essential marker missing). The average G + C content for the DNA is 58.06%. A total of 3,342 coding DNA sequences (CDSs) were predicted for the chromosome, 4 pseudogenes and 10 miscellaneous RNAs (misc-RNA), 3 rRNA operons, and 54 tRNA genes.
The three 16S rRNA genes of Desulfovibrio sp.86 are identical. Their best BLAST hits (NCBI) were from uncultured Desulfovibrio sp. clones from Microbial Fuel Cells such as MFC63A04 (Genbank accession number : FJ823865; coverage 98%; identity 99,87%) 20 . A phylogenetic tree using the available 16S rRNA of cultivable bacteria confirmed the similarities between Desulfovibrio sp.86 and Desulfovibrio simplex DSM4141 (Genbank 16S rRNA accession number: NR_117110; coverage 99%; identity 99, 22%; genomic sequence unavailable) (Fig. S2) 21 . At the genomic level, Desulfovibrio sp.86 ranks first with Desulfovibrio desulfuricans subsp. desulfuricans str. ATCC 27,774 genome (GenBank:NC_011883). Nevertheless, Desulfovibrio sp.86 shares only 1,408 genes (43%) with D. desulfuricans (over 80% amino acids identity and 80% alignment coverage). In addition, the average nucleotide identities (ANI) between Desulfovibrio sp.86 and sequenced Desulfovibrio genomes are lower than the 95% ANI cut-off value generally accepted for species delineation (Fig. S3) 22 . These results indicate that Desulfovibrio sp.86 is most probably a new species of the Desulfovibrio phylum. The presence of two superoxide dismutases and one catalase gene accounts for its relative oxygen tolerance. As expected, the Desulfovibrio sp.86 genome exhibits an extensive gene complement for sulfur metabolism, encompassing the sulfur respiration pathways with inorganic sources like sulfate, sulfite, bisulfite and tetrathionate as electron acceptors. Organic sulfur sources may be provided by fermentation products of sulfur biomass (sulfoquinovose of sulfoquinovosyl lipids) sulfoacetate, the sulfur non-proteogenic amino acid taurine via sulfoacetaldehyde, or alkanesulfonates 23 .
Desulfovibrio sp.86 degrades chlordecone into known transformation products as well as an unexpected sulfur-containing compound. The chlordecone-degrading ability of the isolated Desulfovibrio sp.86 strain was investigated using GC-MS (Gas Chromatography Mass Spectrometry) and LC-HRMS (Liquid Chromatography High Resolution Mass Spectrometry) techniques in growth conditions successfully applied for Desulfovibrio sp.86 isolation (MMD liquid medium). Na 2 S was used as reductant and an anaerobic N 2 /H 2 (98/2; V/V) atmosphere was applied using a glove box system. These incubation conditions led to complete disappearance of the chlordecone signal in GC-MS and LC-HRMS, and resulted in similar GC-MS and www.nature.com/scientificreports/ LC-HRMS TP profiles as those obtained with Citrobacter sp.86 6 : monohydrochlordecone A1, pentachloroindene B1, tetrachloroindenes B3-B4 and polychloroindenecarboxylic acids C1-C2 and C3-C4 (Fig. 1a,b). In the glove box system, bacterial cultures were performed using 100 mL glass vials with a hydrophobic porous film to enable gas exchange and avoid contamination. This incubation condition was considered as "renewed atmosphere" condition (RA). In this system, the atmosphere was regularly renewed with N 2 /H 2 (98/2; V/V) and the box was not thermostatically controlled so the temperature varied between 25 and 33 °C. To better control growth and degradation conditions, Desulfovibrio sp.86 cultures were placed in MMD medium using 100 mL glass vials, sealed with butyl rubber septa, in an oven at 30 °C. Sealed vials were initially purged with N 2 /H 2 (98/2; V/V) gas to insure anaerobiosis. This incubation condition was considered as "confined atmosphere" conditions (CA). After six-week incubation with chlordecone, the pesticide was no longer detectable by the same analytical techniques. Concomitantly, a single unknown chlorinated compound named F1 appeared. It was detectable only through GC-MS (25.6 min) (Fig. 1c). Like in the previously reported chlordecone degradation cultures 5,6 , the main chlordecone transformation appeared during the stationary phase (Fig. 1e). Since no visible chlorinated peak could be observed in LC-HRMS (Fig. 1d), we based the structural elucidation of F1 on the interpretation of GC-MS data (Fig. 1f). Assuming that the higher isotopic pattern centered at m/z 507.8 contained the molecular ion, we hypothesized two possible neutral formulae for F1, C 10 Cl 10 O 2 H 2 and C 10 Cl 10 SH 2 . In-source fragments were highly similar to those observed in polychlorinated bishomocubane-based structures including chlordecone, hydrochlordecones, chlordecol and mirex 5,6,24 . Indeed, the isotopic patterns at m/z 201.0, 235.9 and 271.9 presumably arose from the known in-source bishomocubane retrocyclodimerization and corresponded to positively charged C 5 -fragments. Comparison with isotopic simulations limited the possibility to the following radical ions: [C 5 Cl 3+n SH 2 ] +· [C 5 Cl 5+n ] +· and [C 5 Cl 3+n O 2 H 2 ] +· , with n = 0,1. The latter bis-oxygenated generic formula turned to be highly unlikely since it required the presence of a gem-diol function or a gem-chlorohydrin moiety on the cyclopentenyl ring that had to resist the harsh GC chromatographic conditions (> 200 °C). Instead, we concluded that a sulfur moiety, i.e. a thiol function, was probably present on the bishomocubane polycycle. We proposed for F1 the structure depicted in Fig. 2a and suggested the name chlordecthiol, the sulfur analogue of chlordecol (chlordecone alcohol).

Synthesis of the chlordecthiol standard and confirmation that transformation product F1 is
identical to chlordecthiol. In order to confirm the structure of F1, the standard chlordecthiol was chemically synthesized and fully characterized. To achieve chemical reductive sulfidation of chlordecone, two steps were needed. The first one consisted in the conversion of the gem-diol function of chlordecone in equilibrium with the corresponding ketone form 20,25 into the sulfur analog, i.e. gem-thiol/thiocarbonyl moiety. To perform this step, phosphorus-containing sulfur reagents are generally applied 26,27 . Here we used phosphorus decasulfide (P 4 S 10 ) also called Berzelius reagent 27,28 according to the protocol of Zaidi and coworkers who successfully synthesized camphorthiol 29 (Fig. 2a). The second step consisted in the reduction of the gem-thiol intermediate using NaBH 4 . After purification, a white solid was obtained in an overall yield of 73% (Fig. 2a). 1D-and 2D-NMR analyses confirmed the chlordecthiol structure: (i) two 1 H signals integrating each for one proton and coupled together (δ 3.78 ppm, d, J = 5.5 Hz and δ 2.01 ppm, d, J = 5.5 Hz), with the one at δ 3.78 ppm correlated to a 13 C signal shifted downfield (δ 55.1 ppm) in HSQC experiment that perfectly account for the CH moiety next to the thiol function and (ii) a total of six visible 13 C signals reflecting the plane of symmetry conserved in the present bishomocubane structure (Fig. 2a, Figs. S19-23). Although in microbial transformation, F1 could not be detected using LC-HRMS tool in the developed conditions, a concentrated sample of the synthetic standard chlordecthiol provided a significant signal. The presence of a sulfur atom and the expected neutral formula C 10 Cl 10 SH 2 were confirmed ( Fig. 2c,d), and the raw formulae C 10 Cl 10 O 2 H 2 was ruled out (Fig. 2b). Lastly, GC-MS analysis demonstrated the perfect match (i.e., the same retention time of 25.6 min and same in-source mass spectra After one-month incubation under CA conditions, 10-monohydrochlordecone A1 was completely transformed into two unknown chlorinated compounds barely separated using GC-MS (retention times: 24.3 min F2 and 24.5 min F3, Fig. 3, Figs. S7-S11). They showed an identical in-source mass spectrum that was highly similar to the F1 fragmentation pattern (Fig. S8). All detected in-source fragments turned to possess one chlorine atom less than their analogues in F1 mass spectrum. Compounds F2 and F3 were thus assumed to be diastereoisomers of 10-monohydrochlordecthiol (Fig. 3c). This was confirmed by the chemical synthesis of the two 10-monohydrochlordecthiol standards from 10-monohydrochlordecone A1 using the procedure previously applied for the synthesis of chlordecthiol F1 (Figs. S24-27). These chemical standards had the same retention times (24.3 and 24.5 min) and the same mass spectra compared to the biological F2 and F3 (Fig. S8). TPs B1, C1-C2 and F1 remained unchanged even after six-month incubation with Desulfovibrio sp.86 under CA conditions.
After six-month incubation under RA conditions, chlordecthiol F1 was partially converted into 10-monohydrochlordecthiols F2 and F3, and two other unknown chlorinated species detected using GC-MS, named F4 (retention time: 17.9 min) and F5 (retention time: 26.6 min) (Fig. 3d, Fig. S12). None of these two compounds gave any detectable signal in LC-HRMS. GC-MS analysis enabled to propose C 10 Cl 6 SH 4 as raw formula for F4 Scientific RepoRtS | (2020) 10:13545 | https://doi.org/10.1038/s41598-020-70124-9 www.nature.com/scientificreports/ Two sulfur-containing inorganic compounds, the reductant Na 2 S and the electron acceptor Na 2 SO 4, were present in the original MMD liquid medium. A first series of experiments in sealed vials with substitution of Na 2 S by other reducing agents, sulfured or not, i.e. cysteine and titanium(III) citrate (TiCi) led to a comparable level of chlordecthiol F1 ( Table 1). Traces of TP B1 were also observed in all experiments, including Na 2 S, the positive control (Table 1).
A second set of experiments was designed using Na 2 S as reducing agent and alternative sulfur-based electron acceptors in sealed vials ( Table 2). Use of 90% 34 S-enriched inorganic sulfate resulted in a chlordecthiol product showing similar 34 S-enrichment level (90% 34 S-F1/10% F1, Fig. S16). GC-MS analysis of culture headspace also revealed the presence of H 2 34 S and H 3 C 34 SH (Fig. S16), thus showing that Desulfovbibrio sp.86 produced H 2 S from sulfate. These gas were detected in the culture headspace in CA conditions, using MMD medium, whereas they were absent from the culture headspace in RA condition that corresponded to the glove box atmosphere (Fig. S17). Absence of sulfate did not inhibit Desulfovibrio sp.86 growth, however it resulted in formation of B1 (Table 2) 5 . The remplacement of sulfate by sulfite, bisulfite or thiosulfate led in all cases to the formation of chlordecthiol F1.
A third series of experiments using jars investigated the effect of the nature of the gas phase in contact with the culture. Between the RA condition using open vials in glove box and CA condition using sealed vials in oven, several parameters differed (atmosphere nature and volume, temperature). Using a jar system including several vials, we selectively study the effect of atmosphere renewal on chlordecone transformation by Desulfovibrio sp.86 in MMD medium. In each case, open vials with hydrophobic porous films were placed in jars initially purged with a selected gas. All the jars were incubated at 30 °C. Some of them were flushed several times, to mimic the glove box system, whereas others were kept closed. Jars 1-4 contained two identical open vials filled with MMD, chlordecone and Desulfovibrio sp.86 inoculum, and one negative abiotic control. Among them, two jars were flushed twice a week, jar 1 with (N 2 /H 2 (98/2; V/V)), jar 2 with N 2 (Fig. 4a), whereas jars 3 and 4, initially purged with N 2 and (N 2 /H 2 (98/2; V/V)) respectively were left unflushed (Fig. 4b).
In parallel to the jars, two sealed vials filled with MMD without sulfate, chlordecone and Desulfovibrio sp.86 were flushed with extemporarily synthesized H 2 S (Fig. S4).
After two-month incubation at 30 °C, TP B1 was detected in vials placed in the flushed jars 1 and 2 (Table 3a), whereas TP F1 was only found in vials placed in unflushed jars 3 and 4 (Table 3b). No TP were present in the abiotic controls. In jars 5 and 6, F1 was present in both sulfate and sulfate-free Desulfovibrio sp.86 open cultures (Table 3c). In the same way, Desulfovibrio sp.86 sulfate-free cultures, flushed with H 2 S showed significant levels of F1 TP, whereas no transformation was observed in the negative control (Table 3d).
An additional chemical experiment was carried out to assess the influence of H 2 S on chlordecone transformation. Classic chemical protocol enabling B and C TPs formation, was applied (chlordecone, titanium citrate, vitamin B 12 , water). Under N 2 atmosphere, B and C TPs were obtained, however under H 2 S atmosphere, only A1 was produced (Fig. S18).  www.nature.com/scientificreports/ These results clearly show that the formation of chlordecthiol F1 requires a closed incubation system with a reducing atmosphere. However H 2 as the initial gas atmosphere is not mandatory whereas the presence of H 2 S is required. Chemically, the presence of H 2 S prevents from ring-opening dechlorination process.
Environmental relevance of chlordecone reductive sulfidation. We reinvestigated our previous collection of the Martinique Island chlordecone-contaminated environmental samples (eight soils and two bed sediments) in which various levels of chlordecone TPs had been reported 6 . Among the novel sulfur-containing TPs herein reported, we found appreciable levels of F1 in the two bed sediments samples (927 and 928) at a concentration estimated around 50 μg/kg and 20 μg/kg of wet sediment, respectively (Table S1). In parallel, bacterial population diversity data issued from metabarcoding analysis of these samples (Fig. 5, Fig. S5) were processed to recover a hierarchical clustering of environmental samples according to their taxonomic composition 6 . It was noticed that sulfate-reducing bacteria were much more present in bed sediments than in the other compartments, as previously reported [30][31][32] .

Discussion
To date, the mechanism of chlordecone microbial transformation by ring-opening dechlorination process has not been reported, nor has any associated enzyme activity been described. Bacteria from the genus Desulfovibrio are indeed known to be part of consortia able to dechlorinate molecules like aromatic hydrocarbons [32][33][34][35] . Desulfovibrio sp.86 is able to produce the same TPs of chlordecone as previously described 5,6 . However, the only other complete genome of a chlordecone degrading organism is from a Citrobacter species. The large divergence between these two bacterial species at the level of genus, lifestyle and metabolism precludes a straightforward differential genomic approach searching for common genes to understand ring-opening dechlorination process at the level of the pathway. Even if a parsimonious hypothesis could invoke the same mechanism for chlordecone ring-opening dechlorination process in both species, additional isolated species and genomes are welcomed to give clues to this question.
In this work, we also report on the identification, in the presence of Desulfovibrio sp.86, of another chlordecone transformation product issued from a reduction process, the sulfured molecule which we named chlordecthiol and which had not been described so far. The sulfur metabolism of bacteria from the genus Desulfovibrio is versatile, and encompasses the use of various inorganic sulfur sources that eventually can be reduced to H 2 S. The Desulfovibrio sp.86 gene content is consistent with the production of sulfide. In particular, under the experimental incubation conditions we used, reduction of sulfate from the mineral medium can be achieved by the sulfate and sulfite reductase genes. www.nature.com/scientificreports/ The reductive sulfidation was also observed with the monohydrochlordecone A1 and then could be extrapolated to other hydrochlordecones. Remarkably, putting F1 in renewed atmosphere conditions essentially leads to sulfured analogs of F1 (F2 and F3), and the methylated derivative F5. This process generates a greater diversity of chlordecone TPs and extends the list of chlordecone derivatives that could be present in the environment. According to LC-HRMS analyses, F1 appeared less polar than chlordecone and chlordecol (Table S2). Indeed, its physico-chemical properties are likely to be different from those of chlordecone due to the replacement of the gem-diol fraction (chlordecone) by the thiol function (F1). The fate of F1 in the environment is therefore  www.nature.com/scientificreports/   www.nature.com/scientificreports/ likely to be different in terms of biodegradability, stability, mobility to other environmental compartments and/ or biomagnification along the food chain. Sulfate reducing bacteria are frequently encountered in anaerobic environments like sediments where Fe(III) and sulfate dominate the pool of electron acceptor 34 . Thus, the environmental relevance of chlordecthiol detection that we report herein in chlordecone-contaminated sediments from the French West Indies may originate from microbial processes. Moreover, our work in laboratory condition indicates that at least some TPs of chlordecone are also prone to reductive sulfidation (hydrochlordecones). Even if at this stage a large set of questions related to their appearance, persistence, trophic food-chain transfer and toxicological profile can only be envisioned, our results open a new window on the secondary contamination by chlordecone TPs as well as possible clues to its natural elimination.
Additionally, the change in the TP profile observed in our experiments with Desulfovibrio sp.86 under various conditions of sulfur donor and atmosphere (confined or renewed in anaerobic conditions) could indicate biochemical pathways blockade and/or promotion triggered by H 2 S accumulation, leading to re-routing of chlordecone into this newly described reductive sulfidation instead of the previously reported ring-opening dechlorination TPs. At this stage, it is worth mentioning that all transformations of chlordecone with the consortium 86 had been conducted in open vials conditions, i.e. renewed atmosphere conditions. In a study related to dechlorination of trichlorofluoromethane by sulfate reducing bacteria, sulfide was proved to be inhibitory to the dehalogenation process 35 .
Some papers mentioned that H 2 S or HScan bind to metal centers and have an inhibitor effect on zinc and cobalt γ-class enzymes 36,37 . It was suggested that corrinoids could play a role in chlordecone ring-opening dechlorination conditions 5 and broadly, vitamin B 12 , known to be a cofactor of reductive dehalogenases, can also intervene in the reductive dehalogenation processes as protein-free corrinoid [38][39][40] , in both biotic 41 and abiotic settings 6,42 . In this work, we showed that vitamin B 12 combined with titanium citrate and chlordecone under N 2 atmosphere gave B and C families, whereas the same combination under H 2 S atmosphere only led to A1 formation (Fig. S18). The weakness of the bond strength of cobalt with H 2 O inducing a spontaneous reductive substitution of aquo − cobalamin(III) (Cbl(III) − H 2 O) by sulfide − cobalamin(III) (Cbl(III) − S 2− ) in the presence of S 2− was reported 43 . In the case of corrinoid implication in B and C families formation, it could be assessed that the presence of H 2 S could also be inhibitory and thus favored F1 by another mechanism.
Reductive sulfidation of ketones and aldehydes are possible using a large variety of chemical protocols, very distant from physiological conditions 27 . Several C-S bond formation are described, some mechanisms involve a glutathion-ester intermediate or a cysteine precursor to form thiol derivatives via nucleophilic reactions 44,45 , others use radical SAM enzymes 45 . Several enzymes known to catalyze sulfur addition reactions could take part in the reductive sulfidation of chlordecone. Among them thiolase, sulfurtransferase, sulfotransferase, sulfhydrylase and numerous Fe-S proteins are predicted in Desulfovibrio sp.86 genome. However, at this stage there is no clue enabling to suggest a mechanism and even to confirm an enzymatic activity toward chlordecone or TPs that were www.nature.com/scientificreports/ shown to undergo reductive sulfidation in Desulfovibrio sp.86 cultures. Further experiments will be necessary to consider if any of these processes are indeed operating with Desulfovibrio sp.86. Desulfovibrio sp.86 is the first sulfate-reducing bacterium known to date that is able to transform chlordecone, and only the second isolated and sequenced bacterial species coping with this pesticide. In this work, we reported for the first time on microbial reductive sulfidation, which occured on substrates like chlordecone or hydrochlordecone. This raises the question of whether the microbial sulfidation process is restricted or not to these peculiar substrates, and whether it can be ascribed to the intrinsic sulfur-reducing ability of Desulfovibrio sp.86. In this respect, this new microbial strain, cultivated under our experimental conditions that promote reductive sulfidation, could provide a framework to investigate further this process with other natural or synthetic compounds. Finally, this study demonstrates that the way anaerobiosis is managed in the laboratory (e.g., renewed atmosphere in the glove box or atmosphere confined in sealed vials) is of critical importance for microbiological cultures as illustrated here by the change in the transformation pathways of chlordecone.
HS-GC-MS (Head Space Gas Chromatography Mass Spectrometry) analyses were performed on a Thermo Fisher Trace 1300 coupled to ISQ 7000 VPI single quadrupole mass spectrometer. The instrument was equipped with a 30 m × 0.25 mm × 0.25 µm DB-624-UI column (Agilent J&W), a split/splitless injector and an automatic sampler TriPlus RSH coupled to a HeadSpace tool. For mass spectrometry (MS) analyses, the following standard working conditions were applied: electronic impact ionization, positive mode detection, ion source at 220 °C, detector voltage 70 eV, full scan mode m/z 33-300 (scan time 0.20 s). Injection and transfer line temperatures were set up at 200 °C, respectively 280 °C. 1 mL was injected each time with a filling speed of 10 min/mL, an injection speed of 10 mL/min and a penetration speed of 10 mL/s. For H 2 S detection, vials were incubated for 1 min at 40 °C and sampled with a syringe at 40 °C. The split mode was applied (30 °C, flow rate at 16.7 mL/min, with a ratio of 33.4). The carrier gas was helium at 0.5 mL/ min for 1 min followed by a gradient of 0.05 mL/min until reaching 1 mL/min (hold time 9 min). GC program was an isocratic at 24 °C for 20 min.
LC-HRMS analyses were carried out using a Dionex Ultimate 3000 LC system coupled to an LTQ-Orbitrap Elite mass spectrometer (Thermo Fisher Scientific) fitted with a heated electrospray ionization source (HESI) operating in negative ionization mode. Voltage optimization was described elsewhere 5 . Chromatographic separation was achieved using a Thermo Fisher Syncronis C18 column (50 mm length, 2.1 mm inner diameter, 1.7 µm particle size) and carried out at 30 °C with a flow rate of 0.5 mL/min using NH 4 OAc buffer (10 mM; pH 7 adjusted with NH 4 OH) as solvent A and MeCN as solvent B. The gradient started at 20% B for 1 min, followed by a linear gradient at 100% B for 7 min and remained 2 min at 100% B. The system returned to the initial solvent composition in 2 min and was re-equilibrated under these conditions for 2 min.
Organochlorine extraction for microbiological culture monitoring. After homogenization of the liquid culture, 500 μL of the turbid solution were collected and extracted twice using 250 μL isooctane. The combined organic layers were then analysed through GC-MS analysis. When HS-GC-MS was required, 700 μL of cultures were sampled and put into a Chromacol 10-HSV vial of 10 mL (Agilent). For LC-HRMS analysis, 50 μL of the turbid solution were collected, mixed with 200 μL of acetonitrile/NH 4 OAc buffer (10 mM; pH 7 adjusted with NH 4 OH) (V/V, ¼) and filtered on 0.22 μm filters.
Genomic DNA extraction, sequencing and analysis. DNA extraction was performed as described previously 5 .
Genome sequencing was performed mixing short and long reads (Illumina technology and Oxford Nanopore technology (ONT)). The Illumina and ONT data, around 280-and 20-fold coverage, respectively, were assembled using Unicycler (version: v0.4.6) (https ://cab.spbu.ru/softw are/spade s/) as described 46 . The annotation and research of putative orthologous relations between genomes defined as gene couples satisfying the bi-directional best hit (BBH) criterion or a blastP alignment threshold with a minimum of 35% sequence identity on 80% of the length of the smallest protein was performed using the Microscope platform 22  www.nature.com/scientificreports/ ously described 5 supplemented with lactate as carbon source (20 mM), yeast extract (1 g/L), Na 2 SO 4 (7 mM) and Na 2 S as reducing agent (0.4 g/L). In the glove box, culture vials were open, equipped with a hydrophobic porous film (VWR, 114 μm thick, medical grade).
Desulfovibrio sp.86 cultures were inoculated at the onset of an experiment with 0.5 mL pre-culture pre-grown in oven for 24 h.
During degradation reactions, the MMD medium was supplemented with 40 mg/L of chlordecone (from a solution of chlordecone at 200 mg/mL in dimethylformamide). Each experiment, tested condition or chlordecone degradation was done in duplicate, plus a negative control (without bacteria) in 100 mL glass serum vials containing 50 mL of mineral medium.
Anoxic microbial degradation of chlordecone with different reducing agents. Desulfovibrio sp.86 cultures were incubated in CA conditions with chlordecone. They were carried out in MMD medium. And when indicated the reducing agent (Na 2 S, 0.4 g/L) was replaced by cysteine (1 g/L) or titanium citrate (0.38 mM).

Anoxic microbial degradation of chlordecone with different electron acceptors. Desulfovibrio
sp.86 cultures were incubated in CA conditions with chlordecone. They were carried out in MMD medium. And when indicated the electron acceptor (Na 2 SO 4 , 7 mM) was replaced by Na 2 34 SO 4 , Na 2 SO 3 , NaHSO 3 or Na 2 S 2 O 3 (7 mM). Na 2 34 SO 4 was 34 S enriched at 90%. For cultures containing Na 2 SO 3 or Na 2 34 SO 4 , the gas volume contained in the vial was analysed after two-month incubation by HS-GC-MS analysis. Environmental samples extraction and analysis. Environmental samples origin and treatment were already described elsewhere 6 . They were collected on Martinique Island. Soil (2 andosol, 2 nitisol, 2 ferralsol soils and 2 ashes and 2 pumice stones) from the vicinity of the "Montagne Pelée" volcano and 2 bed sediment samples from Galion bay were taken from the 0-30 cm layer and conserved in glass boxes.

Gas
Dual chemical analysis of these samples (using GC-MS and LC-MS analysis) was already published elsewhere but without searching and taking into account the possible presence of sulfured TPs 6 . Samples were re-analysed and re-extracted according to the previously reported procedure 6 , to focus on sulfured chlordecone TPs. Each sample was processed in duplicate. For soils and sediments (10 g), 50 mL of milliQ water was added, followed by acidification to pH 1 with HCl (1 M) and vortexing. After decanting, the supernatant was extracted with DCM (10 × 50 mL) and the pellet washed twice with DCM (50 mL). Organic layers were pooled, concentrated in vacuo, and analyzed in duplicate injections by GC-MS (in dichloromethane). Soil sample from Martinique taken at a location known not to be contaminated with chlordecone (Nitisol_neg) was used as negative control, and was prepared and treated as mentioned above. GC-MS calibration curve was done using pure F1 compound in dichloromethane.
Soil biodiversity analysis was done on the previously described soil and bed sediment samples.The extraction protocol and the 16S rRNA gene pyrosequencing and analysis is described in Supplementary Information. Putative sulfate-reducing bacteria were assessed according to 16S rRNA sequences and according to the known sulfate-reducing bacteria already reported 17 . The list of families found in this analysis is given in Fig. S5. The dendrogram generated from hierarchical clustering of environmental samples was done using R package pvclust v. 1.3-2 47 .
Chemical synthesis of F1 TP. Phosphorus pentasulfide (400 mg, 9.0 10 -4 mol, 4.5 eq) was added to a solution of chlordecone (100 mg, 2.0 10 -4 , 1 eq) in pyridine (20 mL). The reaction mixture was stirred under N 2 , at reflux for 12 h. After cooling, 100 mL of pentane was added and the organic layer was successively washed with hydrochloric acid (2 M, 3 × 50 mL), distilled water and brine. Organic phase was dried over MgSO 4 , and concentrated under reduced pressure to give rise to an orange crude solid. To the crude solid, NaBH 4 (22 mg, 6.0 10 -4 mol, 3 eq) was added in 2-propanol (10 mL). The reaction mixture was stirred under N 2 , at room temperature, for 24 h. It was evaporated under reduced pressure and 10 mL of 5% H 2 SO 4 was added. The acidic