Chlordecone was obtained from Azur Isotopes (purity 98%). Chemical products used for microbiological media, vitamin B₁₂ (>98%), chloro(pyridine)cobaloxime(III), lindane (97%), 1,3-dichlorobenzene (98%), 1,4-dichlorobenzene (99%), chlorobenzene (99.8%), and benzene (99.8%) were obtained from Sigma Aldrich. Titanium (III) citrate was prepared from titanium (III) chloride (>12% in HCl; Sigma Aldrich) and sodium citrate and neutralized with Na₂CO₃ (Chevallier et al., 2019). Dichloromethane (HPLC grade) was obtained from Fisher Chemical.

Knockout mutant strains in Citrobacter sp.86 were constructed using the λ-red recombinase technique as developed for Escherichia coli (Datsenko and Wanner, 2000) with some modifications. Briefly, a PCR product was generated by using primers with 50-nt 5'-extensions that are homologous to regions adjacent to the gene to be inactivated and 20-nt 3'-extremities hybridizing to the antibiotic resistance cassette of the plasmid pKD3 (chloramphenicol resistance cassette) or pKD4 (kanamycin resistance cassette). The amplicon was purified (QIAquick PCR Purification Kit, Qiagen) and introduced by transformation in electrocompetent Citrobacter sp.86 cells which harbor the helper plasmid pKD46-Gm (Doublet et al., 2008; kindly provided by Benoit Doublet). The thermosensitive plasmid pKD46-Gm contains the Red recombinase genes located under an L-arabinose inducible promoter and the gentamicin resistance cassette. Recombinant cells were selected on LB-agar plates containing the appropriate antibiotic (kanamycin or chloramphenicol, 50 μg/ml). The effective replacement of the target gene with the resistance cassette was verified by PCR-amplifying the targeted chromosomal locus of wild-type strain and of the recombinant candidates using specific primers located 200 nt upstream and downstream of the target gene and analyzing the PCR products by electrophoresis on agarose gels. The primer sequences used in this study are given in Tables S1, S2, and the designation and genotype of all bacterial strains as well as plasmids used in this study are given in Tables S3, S4.

Citrobacter sp.86 strains were kept frozen at −80°C as stock glycerol (15%). Before starting phenotyping or degradation experiments, the bacteria were first grown aerobically at 37°C on LB plates with 100 μg/ml carbenicillin (the Citrobacter sp.86 genome encodes a beta-lactamase), and isolated colonies were used as inoculum.

For analysis of growth requirements (phenotyping), the Citrobacter sp.86 wild-type and mutant colonies issued from LB agar plates were then spread onto agar plates containing mineral medium MM previously described (Chaussonnerie et al., 2016) but without vitamin B₁₂, supplemented with glucose (20 mM) or pyruvate (40 mM). When indicated, vitamin B₁₂ and L-methionine were added at a final concentration of 0.3 μM or 0.3 mM, respectively. Plates were incubated aerobically at 37°C. For anaerobic assays, colonies issued from aerobic MM plates supplemented with methionine but without vitamin B₁₂ were spread on MM plates containing Na₂S (0.4 g/L), glucose (20 mM), or pyruvate (40 mM), with or without vitamin B₁₂, and incubated at room temperature (rt) in a glove box (Unilab mBraun), under an N₂/H₂ (98/2; V/V) atmosphere.

Anoxic microbial incubations (degradation experiments) were performed in the glove box in daylight. Microbial liquid cultures were carried out in the mineral medium MM, but without vitamin B₁₂, supplemented with pyruvate (40 mM) as carbon source, 2 g/L yeast extract, and 2 g/L tryptone. This complemented mineral medium was named MMpyt. The reductant was Na₂S (0.4 g/L), and 0.1% of resazurin was added as an indicator of anaerobiosis. When indicated, it was adjusted at a pH differing from the standard (pH 7.0) by varying the proportions of the buffering system (KH₂PO₄/K₂HPO₄). Chlordecone and lindane were solubilized in dimethylformamide to a 200 mg/ml stock solution and used at 40 μg/ml or 20 μg/ml, respectively.

After growth on LB plates, a colony of each Citrobacter sp.86 strain resuspended in 50 μl NaCl 0.8% was transferred in the anaerobic glove box (Figure S2) and used for initial inoculation of a 2 ml culture in MMpyt with 100 μg/ml carbenicillin (culture C1). After 24 h, 50 μl of C1 was used to inoculate a 5 ml culture in MMpyt with 100 μg/ml carbenicillin and 10 μg/ml chlordecone or 10 μg/ml lindane (culture C2). After 24 h, C2 was finally used to inoculate a 50 ml culture in MMpyt and 40 μg/ml chlordecone or 20 μg/ml lindane (culture C3, contained in 100 ml glass serum vials), which was incubated for 4 months and 28 days for chlordecone and lindane, respectively. Each degradation experiment was done in duplicate. All experiments were monitored over time using GC-MS techniques. A negative control vial (without bacteria) was added to each tested condition.

Chlordecone monitoring was performed using GC-MS. After homogenization of the liquid cultures, 500 μl were collected and extracted twice using 250 μl isooctane. The combined organic layers were then analyzed through GC-MS analysis via liquid injection.

For lindane monitoring Headspace GC-MS (HS-GC-MS) was required. In this case, 700 μl of culture were sampled and put into a Chromacol 10-HSV vial of 10 ml (Agilent). The headspace gas was then analyzed through the Headspace tool (see below).

Gas Chromatography coupled to Mass Spectrometry analyses were used for chlordecone degradation monitoring and were carried out using a Thermo Fisher Focus GC coupled to a single-quadrupole mass spectrometer (Thermo Fisher DSQ II). The instrument was equipped with a non-polar 30 m × 0.25 mm × 0.25 μm DB-5MS column (Agilent J&W) and a split/splitless injector. Ionization conditions and the GC program have been described elsewhere (Chevallier et al., 2019).

Gas Chromatography Mass Spectrometry coupled with a Headspace trap (HS-GC-MS) analyses were used for lindane degradation monitoring and were performed on a Thermo Fisher Trace 1300 coupled to an 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 MS analyses, the following standard working conditions were applied: electronic impact ionization, positive mode detection, ion source at 220°C, detector voltage 70 eV, and full scan mode m/z 33–300 (scan time 0.20 s). Injection and transfer line temperatures were set up at 200 and 280°C, respectively. Monitoring vials were incubated for 5 min at 50°C and sampled with a syringe at 50°C. One milliliter of the headspace gas was injected each time at a filling speed of 10 ml/min, an injection speed of 10 ml/min, and a penetration speed of 10 ml/s. The splitless injection mode was applied at 150°C. Carrier gas was helium at a constant flow rate of 0.5 ml/min. The GC program started at 30°C (hold time 6 min), continued with 15°C/min to 130°C (hold time 0.5 min), followed by 7°C min⁻¹ to 250°C (hold time 10 min).

According to the protocol described elsewhere (Chevallier et al., 2019), to a solution of chlordecone (5.0 mg, 9.9 10⁻⁶mol, and 1 eq.) and vitamin B₁₂ (4.1 mg, 3.0 10⁻⁶mol, and 0.3 eq.) or cobaloxime (1.2 mg, 3.0 10⁻⁶mol, and 0.3 eq.) in degassed water (30 ml) was added titanium (III) citrate basified to pH 12 with NaOH (3 M; 5 ml, 3.3 10⁻⁴mol, and 32 eq.). The reaction mixtures were stirred under N₂ atmosphere at rt for 2 h; and monitored by GC-MS.

Since most genes predicted to operate in a cobalamin biosynthesis pathway are clustered (Figure 2A), we chose to construct a test-set of mutant strains by targeted deletion of genomic coding regions encompassing multiple genes. In fact, these multiple gene deletions would certainly result in a defective cobalamin. Three mutant types were designed affecting different steps of the cobalamin molecule construction: (i) the insertion of the cobalt atom into the corrin ring and corrin core functionalization, (ii) the corrin core functionalization and nucleotide loop biosynthesis, and (iii) the nucleotide loop biosynthesis including the lower ligand insertion. They were respectively obtained by deleting the following gene combinations: cbiHJK, cobQUS, and cobST. In addition, based on preliminary results obtained for the Citrobacter sp. 86 ΔcbiHJK strain, a fourth mutant type cbiK affecting the cobalt insertion was constructed (Figure 2B). Detailed reactions catalyzed by proteins encoded by these genes are shown in Figure S3.

There are two methionine synthases in Citrobacter sp.86, one cobalamin-independent encoded by the gene metE and the other cobalamin-dependent, encoded by the gene metH. On a ΔmetE background, a deletion that would impact the biosynthesis of cobalamin in Citrobacter sp.86 would have an effect on its viability. In a first step, we decided to test our constructions on this background.

Prototrophy of a metE knockout mutant strain in mineral medium and in the absence of any nutritional supplement other than a carbon source (glucose or pyruvate) was used as a physiological phenotypic indicator of the cellular cobalamin production. In the four ΔmetE and Δcob/cbi double mutant strains, we hypothesized that rescue of methionine auxotrophy would be achieved by exogenous cobalamin.

The growth phenotype of all strains was tested on solid rich or mineral medium under both aerobic and anaerobic conditions. Independently, all genotypes were also checked by PCR (Figure S4A).

As shown in Table S5, no growth defects were observed for any strain on LB rich medium, but growth phenotypes were more contrasted on mineral medium (Table S5; Figures S4B,C). The Citrobacter sp.86 ΔmetE strain did not grow under aerobic conditions and this defect was alleviated by supplying exogenous cobalamin, which is required as a cofactor for MetH. In contrast, it grew under anaerobic conditions in the absence of any nutritional supplement indicating successful endogenous cobalamin biosynthesis. So, we concluded that Citrobacter sp.86 was unable to synthesize cobalamin under aerobic conditions. These results were consistent with the annotation of the genes involved in the anaerobic cobalamin pathway. Similarly, the enteric bacterium Salmonella typhimurium also synthesizes cobalamin de novo only under anaerobic growth conditions (Jeter et al., 1984).

The four Citrobacter sp.86 double mutant (ΔmetE and Δcob/cbi) strains were unable to grow on mineral medium under anaerobiosis, indicating that incomplete corrinoids cannot functionally replace the cobalamin cofactor for MetH functionality. Under these conditions, exogenous cobalamin fulfills this requirement. These results confirm the impairment of the cobalamin biosynthetic pathway in all cob/cbi knockout mutant strains made in this study.

In a second step, we also constructed four Δcob or Δcbi Citrobacter sp.86 knockout mutant strains to test for their ability to degrade chlordecone and lindane compounds. Based on the results of genotype verification by PCR and the physiological consequences of the cobalamin biosynthesis pathway disruption described above, these strains are impaired in the cobalamin synthesis pathway. None of these deletions impacts growth of Citrobacter sp.86 on mineral medium in the absence of cobalamin (Table S5; Figure S5; Supplementary Material), indicating that production of incomplete non-functional corrinoids was not harmful for the bacteria. Incubation of all these strains with chlordecone or lindane did not modify their growth phenotype (Figure S5; Supplementary Material).

The wild-type Citrobacter sp.86 and the test-set of cobalamin biosynthesis pathway mutant strains were incubated with chlordecone in MMpyt medium. Under these laboratory conditions, the wild-type strain reached the stationary-phase after 10 h. In reported bacterial transformations of chlordecone (Chevallier et al., 2019), the appearance of A and B TP families (monitored using GC-MS) systematically went along with C family (detected using LC-MS) while chlordecone disappearance was completed after several months. Here, we only focused on the detection of A and B families using GC-MS analysis. Transformation profiles of chlordecone in cultures were monitored regularly by GC-MS over a 4-month period. Interestingly, Citrobacter sp.86 and the mutant strains ΔcobQUS and ΔcobST showed the same chlordecone transformation profile (Figure 3; Supplementary Material): upon chlordecone incubation, trace amounts of B1 were detected on the 7th day. After 21 days, B1 was the main chlordecone TP detected using GC-MS analysis, as expected (Chaussonnerie et al., 2016; Chevallier et al., 2018, 2019). The transformation did not occur while the cells were actively growing but rather mostly after entry into the lysis phase (Figure S5). On the other hand, Citrobacter sp.86 ΔcbiHJK and ΔcbiK did not transform chlordecone even after an additional extensive period of incubation (a total period of 201 days, Supplementary Material), just as in the negative control without bacteria. In the light of these results, it can be assumed that suppression of the corrinoid lower ligand or modifications of the corrin core patterns do not affect the ability of Citrobacter sp.86 to degrade chlordecone. In contrast, inactivating the cobalt insertion step prevents chlordecone degradation, confirming the involvement of cobalamin or other corrinoids in chlordecone transformation. However, the corrinoids, which are functional for the chlordecone degradation process, are not necessarily complete cobalamin in contrast to the functional cofactor involved in enzymatic reactions. All these results are consistent with a transformation of chlordecone by a non-enzymatic process involving corrinoids with an inserted cobalt atom. So far, no mechanism can assess how cobalamin would react with chlordecone. However, assuming that chlordecone transformation would be the result of the action of corrinoids released by Citrobacter sp.86, the formation of a Co-C bond linking the bishomocubane cage to the corrinoid could be inferred. It is likely that a Co(I) oxidation state would be needed to enable the chlordecone attack (Schrauzer and Katz, 1978). In 1978, Schrauzer and Katz apparently detected [Co]-C₃Cl₃H₂ fragments within the abiotic reaction mixture of chlordecone with vitamin B₁₂, supporting this assumption.

In addition, it is known that several chemical transformations of chlordecone involving vitamin B₁₂ reduced under a Co(I) oxidation state by strong reducing agents generate the same TP profile (Schrauzer and Katz, 1978; Ranguin et al., 2017; Chevallier et al., 2018, 2019). In our hands, use of chloro(pyridine)cobaloxime, known to be a good model of vitamin B₁₂ (Schrauzer and Katz, 1978; Terán et al., 2018; Pizarro et al., 2019) also led to the same diversity of chlordecone TPs in presence of a strong reducing agent (Figure S6). It supports the hypothesis that several cobalt complexes containing a corrin ring could afford the same chlordecone TPs, once reduced.

The present results show that microbial degradation of chlordecone mediated by Citrobacter sp.86 is clearly correlated to the production of cobalamin derivatives. Chemical degradation using vitamin B₁₂ or cobaloxime in aqueous solution also leads to the same TPs (Schrauzer and Katz, 1978; Jablonski et al., 1996; Ranguin et al., 2017; Chevallier et al., 2018, 2019). Taken together, these observations suggest that microbial and chemical degradations share strong similarities in their mechanistic pathway. These conclusions are apparently not supported by some of our previous findings obtained from carbon specific isotope analysis (CSIA). Indeed, the carbon isotopic enrichment factors observed for microbiological and vitamin B₁₂-mediated degradations differed significantly, suggesting two distinct mechanisms (Chevallier et al., 2018). However, the strong pH variation (pH 7 for microbiological cultures and pH 12 for chemical conditions) could also explain the difference in the ¹³C isotopic signatures. For instance, Heckel and co-workers observed that a change from acidic to basic conditions led to a switch in dechlorination mechanisms during the chemical degradation of trichloroethene mediated by vitamin B₁₂ reduced under a Co(I) state (Heckel et al., 2018).

Furthermore, it is possible that biologically produced corrinoids could also be responsible for chlordecone degradation performed by anaerobic bacteria and archaea described in other studies. Thereby, the chlordecone degradation by the methanogen M. thermophila seemed to be mediated by corrinoids (Jablonski et al., 1996). Interestingly, consortia 86 and 92 able to degrade chlordecone contained, among others, Desulfovibrio, Pleomorphomonas, and Sporomusa species, possessing cobalamin biosynthesis genes (Chaussonnerie et al., 2016). Among these bacteria, Desulfovibrio sp.86 was isolated and showed to degrade chlordecone along with the same TPs profile as observed in presence of Citrobacter sp.86 (Della-Negra et al., 2020). In the same way, in microcosms amended with chlordecone, no known obligate organohalide respiring bacteria were observed, whereas enrichment with Desulfovibrio, Sporomusa, and Geobacter species or methanogens were noticed (Lomheim et al., 2020). Still, in this case, these bacteria and archaea could be corrinoid-producers.

Lindane (γ-hexachlorocyclohexane) dechlorination by C. freundii under anaerobic conditions has been already described (Jagnow et al., 1977). Remarkably, as it is the case for the degradation of chlordecone by Citrobacter sp.86, it was noticed that dechlorination of lindane may not be related to C. freundii growth. In this context, we tested whether Citrobacter sp.86, which is a different species than C. freundii, was intrinsically able to dechlorinate lindane and if this dechlorination was mediated by cobalamin or corrinoids.

Under the same incubation conditions as for chlordecone, in another set of experiments, lindane did not impede the growth of Citrobacter sp.86 (Figure S5). Lindane degradation was observed with the concomitant formation of benzene and chlorobenzene as major TPs, (Figure 4). Trace level of γ-tetrachlorocyclohexene was also detected during lindane degradation operated by Citrobacter sp.86 (Supplementary Material). After 28 days, no more lindane in the culture was detected by HS-GC-MS. In contrast, after 42-day incubation, chlordecone was still detected by GC-MS analysis (Figure 3). The higher solubility in water of lindane (7–9 mg/L at 25°C and pH 7, Saley et al., 1982) compared to chlordecone (1–2 mg/L at 25°C and pH 7, Dawson et al., 1979) may account for its higher degradability (at the same molar ratio) by the Citrobacter sp.86 cultures.

Chlorobenzene, benzene, and/or γ-tetrachlorocyclohexene have already been described as TPs observed in anaerobic degradation of lindane by C. freundii, sulfate-reducing bacteria including D. gigas, D. multivorans, and also in bacterial consortia enriched in Pelobacter (Jagnow et al., 1977; Boyle et al., 1999; Badea et al., 2009; Qiao et al., 2020). It is likely that these TPs appeared by a two-step process involving two sequential dichloroelimination reactions, followed by another dichloroelimination to produce benzene or a hydrodechlorination to generate chlorobenzene (Zhang et al., 2014; Qiao et al., 2020; Figure S1B). After C. freundii (Jagnow et al., 1977), Citrobacter sp.86 is the second species of this genus reported to degrade lindane, and this degradation most likely occurs in the same way as in other anaerobic bacteria. These bacteria dechlorinate lindane cometabolically, with for instance glucose, lactate, or pyruvate as suitable carbon sources.

We also observed that Citrobacter sp.86 ΔcobQUS degraded lindane just like the wild-type (Figure 4B). In contrast, Citrobacter sp.86 ΔcbiHJK and Citrobacter sp.86 ΔcbiK were unable to quantitatively dechlorinate lindane (Figures 4C,D). Low levels of chlorobenzene and 1,3-dichlorobenzene as well as trace amount of 1,3,4,5,6-pentachlorocyclohex-1-ene were observed, but GC-MS peak areas were insignificant compared to the Citrobacter sp.86 wild-type strain and ΔcobQUS mutant strain. In comparison, abiotic controls showed a significant formation of 1,3,4,5,6-pentachlorocyclohex-1-ene while lindane remained definitely predominant after 28 days. This TP has been previously reported during the photolysis of lindane (Zaleska et al., 2000). As all degradation experiments were performed in daylight, we assumed that an additional slow photodegradation process was taking place. The rate of photodegradation would depend on the prevalence of other competing degradation pathways.

An additional abiotic pH-dependence study of lindane dechlorination showed that 1,3,4,5,6-pentachlorocyclohex-1-ene, 1,3-dichlorobenzene, and chlorobenzenewere spontaneously produced, albeit at low rate, when the pH of the incubation medium was set at more basic values (pH ≥ 8), as it has already been described (Figure 5; Hiskia et al., 1997). The possible photodegradadation process and the pH-dependent degradation phenomenon could also explain the detection of 1,3-dichlorobenzene and chlorobenzene in mutants Citrobacter sp.86 ΔcbiHJK and Citrobacter sp.86 ΔcbiK.

These results show that corrinoids synthesized by Citrobacter sp.86 as well as Citrobacter sp.86 ΔcobQUS are involved in the most prominent lindane degradation pathway. As for chlordecone, a complete cobalamin molecule was not needed for lindane degradation. Also, by abiotic processes, Marks et al. (1989) showed high activity of lindane dechlorination by cobinamides reduced with dithiothreitol. In their work, the authors tested a variety of porphyrins and corrins for catalysis of lindane dehalogenation. They showed the importance of the cobalt ion in the tetrapyrrole ring. Also, dehalogenation activity of cobinamides was about 8-fold higher than that of cobalamin. These authors suggested that the lower ligand could sterically hinder the approach of lindane to the cobalt ion at the center of the corrin ring. It could also be that the absence of the lower base coordination results in an increase of the oxido-reduction potential of cobinamide compared to cobalamin, which becomes easier to reduce (Dereven’kov et al., 2016).

Finally, we postulate that the implication of corrinoids, synthesized by Citrobacter sp.86, in lindane dechlorination as shown in this study could be extended not only to C. freundii but also to other bacteria that anaerobically degrade lindane such as D. gigas, D. multivorans, and Clostridium (Macrae et al., 1969; Jagnow et al., 1977; Ohisa and Yamaguchi, 1978; Boyle et al., 1999; Badea et al., 2009; Mehboob et al., 2013). In the same way, corrinoids could be involved in lindane degradation by consortia with the enrichment of Pelobacter within the dehalogenation process (Qiao et al., 2020). All these bacteria (i) are fermenters, (ii) transform lindane co-metabolically, and (iii) seem to be able to produce cobalamin (Roth et al., 1996; Shelton et al., 2019).

Under abiotic conditions, involvement of corrinoids in transformation and dechlorination of compounds including chlordecone and lindane has been well-documented (Schrauzer and Katz, 1978; Marks et al., 1989; Jablonski et al., 1996; Rodríguez-Garrido et al., 2004; Ranguin et al., 2017; Chevallier et al., 2018, 2019; Guo and Chen, 2018; Pizarro et al., 2019). Under biotic conditions, involvement of corrinoids as cofactors in reductive dehalogenases found in organohalide respiring bacteria that transform a variety of organochlorides pollutants was also known (Schubert et al., 2018). Moreover, a dechlorination process was observed in non-organohalide-respiring bacteria and under these conditions, the implication of protein-free corrinoids was already suggested (Jagnow et al., 1977; Badea et al., 2009; Lomheim et al., 2020). This last point was confirmed in that study.

From an environmental point of view, it is highly plausible that corrinoids produced by bacteria could have a significant impact on the natural biodegradation of organochlorine compounds under anaerobic conditions. Pools of cobalamin are produced de novo and subject to complex trade-offs for salvage and remodeling in terrestrial and marine ecosystems (Heal et al., 2017; Lu et al., 2020). In contaminated sites, biostimulation of bacteria producing corrinoids could be a method of depollution (Guo and Chen, 2018). Although, the harsh reductive conditions, which are required for corrinoid-mediated chlordecone and lindane biotransformations in the laboratory may not easily be met in polluted agricultural soils, the presence of degradation products in natural environments (Chevallier et al., 2019) leaves this possibility open.

On a smaller and easier to control scale, for example in reactors, biodegradation could be performed using bacteria producing corrinoids at high yield. Though anaerobic bacteria, for example from the genus Desulfovibrio, are known as producers of corrinoids (Men et al., 2014, 2015), bacteria from the genus Citrobacter like for instance Citrobacter sp.86 could be of great interest in co-metabolic degradations. As we show here, this facultative anaerobic bacterium that displays a versatile anaerobic metabolism involving cobalamin (e.g., synthesis of building blocks, degradation of glycerol, propanediol, ethanolamine, and glutamate…) can be easily manipulated genetically. Tools of the synthetic biology toolbox (Dvořák et al., 2017) like knockin and knockout strategies could therefore be implemented in Citrobacter sp.86 to determine the optimal corrinoids, enhance their production and test for further improvements in lindane, chlordecone, or other organochlorine transformations in the perspective of possible bioremediation improvement.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.