Transformation of ε-HBCD with the Sphingobium Indicum enzymes LinA1, LinA2 and LinATM, a triple mutant of LinA2

Implementation of the LinA1 substrate-binding site to LinA2 by point-directed mutagenesis. Expression of three LinA enzyme variants, LinA1, LinA2 and LinATM, in


Polyhalogenated compounds and persistent organic pollutants
Radical-induced halogenation of various hydrocarbon precursors has been exploited in the past to produce polyhalogenated compounds at large scales. This approach can produce molecular libraries of hundreds of compounds differing in constitution and stereochemistry. For example, the non-specific chlorination of benzene leads to mixtures of nine hexachlorocyclohexane (HCH) stereoisomers. Only one of them, g-HCH (Lindane) is the active insecticide, present at about 8e15% in technical HCH mixtures (Li, 1999;Vijgen et al., 2011). Pure g-HCH was extracted from the crude product in later years while other HCHs remained as waste material (Vijgen et al., 2011). In 2009, a-, band g-HCHs, the most abundant isomers in technical HCH mixtures, were included in the list of persistent organic pollutants (POPs) of the Stockholm Convention (SC) and HCH production and use were banned (UNEP, 2006;2009).
The industrial production of dibromoethane, widely used as a scavenger in leaded gasolines, had to be reduced substantially when noble metal-based three-way catalysts were implemented in vehicles and non-leaded gasolines were introduced (Alaae et al., 2003). This brought about a remarkable shift in bromine use pattern and the advent of brominated flame-retardants (BFRs). Bromination of hydrocarbons is another example of a non-specific halogenation process that leads to mixtures of polybrominated materials. Bromination of 1,5,9-cyclododecatriene isomers yields mixtures consisting of at most 16 HBCD stereoisomers . Fig. S1 (Supplementary material, SM) displays these stereoisomers. Constitutions of seven of the eight isomers found in technical HBCDs, have been deduced from crystal structure analysis 2007a;.
HBCDs were high production volume chemicals too, produced at more than 10 0 000 t/y and used as flame-retardants, mainly in plastic and textile applications (Alaee et al., 2003). As HCHs are, also HBCDs are persistent, toxic, and accumulate along food chains (Tomy et al., 2004;Janak et al., 2005;Law et al., 2005;Kohler et al., 2008;Peck et al., 2008;Vorkamp et al., 2012). As certain isomers of such mixtures are transported over long distances, they reach even remote areas (de Wit et al., 2006). In 2013, after being used for about three decades, HBCDs were congruously added to the POP list of the Stockholm convention (UNEP 2013). Because these chemicals were produced and applied at large scales and are now ubiquitously distributed in the environment (de Wit, 2002;de Wit et al., 2006;Covaci et al., 2006;Marvin et al., 2011), they can be considered as markers of the anthropocene.
Transformations of HCHs are stereoselective. LinA1 preferentially converts the (þ)aeHCHeenantiomer, while LinA2 prefers the (À)a-HCH enantiomer (Suar et al., 2005). Product selectivity differs too and only specific pentachlorocyclohexene stereoisomers are formed (Suar et al., 2005;Shrivastava et al., 2017). Despite these differences, amino acid sequences of LinA1 and LinA2 are quite similar (90%), differing only by 15 out of 156 amino acids ( Fig. 1.) Shrivastava et al. (2017) produced a series of single-, double-and triple-mutants of LinA enzymes by point-directed mutagenesis and identified those amino acids, involved in HCH recognition. The interactions of leucine L96, phenylalanine F113 and threonine T133 seem to be important for the positioning of both a-HCH enantiomers in the active site and with it affect the product selectivity of the enzyme (Fig. 1).
HCHs with diameters of 5e6 Å are relatively small compared to HBCDs (7e9 Å). Nevertheless, we hypothesized that HBCDs might also be suitable substrates for LinA enzymes and exposed 8 of the 16 HBCD stereoisomers (Fig. S1, SM) to LinA2. Evidence was found that both b-HBCD enantiomers are transformed by LinA2 (Heeb et al., 2014(Heeb et al., , 2015. The cleavage of CeBr bonds was found to be selective too and only few of the 32 possible pentabromocyclododecenes (PBCDens) were formed (Fig. S2, SM).
To get more insights into how LinA enzymes recognize and bind HBCD substrates, we expressed LinA1, LinA2 and LinATM, a triple mutant of LinA2, which includes three amino acids of the LinA1 substrate-binding site. We compared the reactivity and selectivity of these LinA variants and determined kinetic parameters and product selectivity of the transformations with ε-HBCD as the substrate. Despite the fact that the three LinA variants all contain the same catalytic dyad, their kinetics and selectivity varied substantially. This confirmed that substrate recognition is controlled by other amino acids that can be exchanged by point-directed mutagenesis. Such exchanges may lead to new LinA variants, with altered substrate and product specificity that may be used for biotransformation of other halogenated POPs.

Cloning, expression and purification of LinA enzymes
The expression of LinA1 and LinA2 has been described before Heeb et al., 2014). Codon-optimized linA1 and linA2 genes were obtained from John Oakeshott (CSIRO, Canberra, Australia). A modified linA2 gene, which carried the desired mutations L96C, F113Y and T133 M, was also cloned. Briefly, codonoptimized (for expression in E. coli) linA1/linA2 genes of S. indicum B90A and genes for the LinA triple mutant L96C, F113Y, T133M, synthesized by GeneArt (Germany), were cloned in a pDEST™17 vector. The plasmid pDEST17 was co-expressed in E. coli BL21-AITM cells with the pGro7 chaperone plasmid through heat shock transformation.
Three E. coli pre-cultures were produced. After addition of inoculum (8 mL), three batches of cloned E. coli bacteria were fermented in lysogeny broth (800 mL), which contained ampicillin and chloramphenicol. Cells were grown over night at 37 C to optical densities at 564 nm (OD 564nm ) of 0.4e0.6. L-arabinose (1.6 g) was added and incubation was proceeded at 30 C until OD 564nm of 1.8e2.0 were reached. Resulting cell material was harvested at 4 C after centrifugation (10 0 000 rpm) and stored at À20 C. Disruption of cells by ultrasonification and isolation of the three LinA enzymes by affinity chromatography were performed following published procedures (Bala et al., 2012), which are described in detail in SM. Three solutions, with enzyme concentrations of 0.91, 2.06 and 0.50 mg mL À1 , were obtained for LinA1, LinA2 and LinATM, respectively, which were further used for exposure experiments.
Mass spectrometric analyses were performed on a triple-stage quadrupole MS (6460 TQ-MS, Agilent) with electrospray ionization in negative ion mode. Nebulizer-and sheath-gas temperatures of 325 and 340 C with flow rates of 8 and 12 L min À1 and needle-, capillary-, fragmentor-, collision-and accelerator-voltages of 1500, 3000, 72, 7 and 6 V were applied. HBCDs and PBCDens were detected in multi-reaction monitoring mode (MRM), selecting the three most prominent chloride-adduct ions [MþCl]and recording transitions to bromide at m/z: 79 and 81.

Molecular docking experiments and graphical representations
Molecular docking experiments were performed with the AutoDock Vina software (Trott and Olson, 2010) as described before (Heeb et al., 2015). Input files in the pdbqt format were used for LinA2, the receptor, and for (þ)band (À)b-HBCDs, the ligands, either in solid state or in activated conformations. Crystal structure data of LinA2 (Okai et al., 2010) was obtained from the protein database (PDB-ID: 3A76). Crystal structure data of b-HBCD enantiomers (Heeb et al., , 2007a can be obtained from CCDC (ID: 617557). Conformational adaptations were achieved with the Chem3D software. Graphical representations and geometrical parameters of enzyme-substrate complexes were obtained with PyMOL.

Structural properties of LinA1, LinA2 and LinATM enzymes
The LinA2 crystal structure (Okai et al., 2010) is available from the protein database at a resolution of 2.24 Å (PDB-ID: 3A76). Fig. 1 displays the primary and secondary structures of the three LinA enzymes expressed in genetically modified E. coli bacteria. Highlighted are those amino acids (one letter code) involved in the catalytic center (magenta), the halide (blue) and the substrate binding sites (red). The catalytic dyad, consisting of aspartate (D25) and histidine (H73), is conserved in these enzymes. More variations are observed for the halide binding sites (Fig. 1, blue), which in case of LinA1 consists of glutamine (Q20), tryptophan (W42) and leucine (L129) and in case of LinA2 of lysine (K20), tryptophan (W42) and arginine (R129). The halide-binding site in LinATM is identical to the one of LinA2, which is positively charged, while the one of LinA1 is not. The secondary structure consists of five aehelices (violet) and six b-sheets (yellow) (Fig. 1).
From previous work on a-HCHs and several single-, double-and triple-mutants of LinA (Shrivastatava et al., 2017) and own considerations, we hypothesized that the amino acids leucine (L96), phenylalanine (F113) and threonine (T133) may also be involved in substrate recognition and binding of HBCDs (Fig. 1, red). By pointdirected mutagenesis, we produced a triple mutant of LinA2, replacing leucine by cysteine (L96C), phenylalanine by tyrosine (F113Y) and threonine by methionine (T133M). These amino acids are located in the antiparallel b-sheet structure (b4, b5, b6) in the inner surface of the cone-like protein (Fig. 1). These mutations are expected to alter the substrate binding specificity, which is different for LinA1 and LinA2. LinA1 preferentially binds (þ)a-HCH, while LinA2 prefers (À)a-HCH (Suar et al., 2005). Fig. 2 displays views of the tertiary and quaternary structures (ribbon form) of LinA2. The tertiary structure of LinA2 is expected to also represent those of LinA1 and LinATM. The structure consists of an extended antiparallel b-sheet, which forms a narrow cone ( Fig. 2A). A long a-helix (Fig. 1, a 1 ) supports the cone structure. A shorter second helix is located at the entrance of the cone, the later has an inner diameter of about 8e11 Å. The catalytic dyad, the halide-and the substrate-binding sites are all located within the cone. In other words, substrates must access the inner space of the cone and products must be released from it to allow a productive catalytic cycle. Fig. 2B and C displays the quaternary structure of a LinA2 homo-trimer. The three cones arrange in a trifold symmetry (C 3v ) with the extended beta-sheets in close contact, while the long a 1 -helices all point in the same direction, forming the tip of the triple cone. Fig. 2D highlights apolar amino acid residues in yellow, indicating that hydrophobic interactions are stabilizing the quaternary structure of the LinA2 trimer.

Stereoselective transformation of HBCDs with LinA2
Sphingobium bacteria expressing LinA enzymes have evolved in HCH-contaminated soils at geographically distinct locations (Pal et al., 2005). HCH and their pentachlorocyclohexene transformation products have diameters of 5e6 Å. We hypothesized that some of the 16 HBCD stereoisomers (Fig. S1, SM), with diameters of 7e9 Å, might also be suitable LinA substrates, binding the active site in productive ways.
We exposed racemic mixtures of a-, band g-HBCDs and the meso forms dand ε-HBCDs to LinA2 and could detect by LC-MS a substantial conversion of band ε-HBCDs and the appearance of two major and one minor PBCDen product. Fig. S3 compares MRM mass traces of chloride adducts [MþCl]of HBCDs and PBCDens. The precursor ions at m/z: 674.7, 676.7 and 678.6 (left) and m/z: 592.7, 594.7 and 596.7 (right) were monitored during fragmentation to bromide ions at m/z: 79 and 81 before (t ¼ 0 h) and after LinA2 exposure (t ¼ 32 h). On a column with a chiral b-cyclodextrin phase, we could separate a-(A), b-(B) and g-HBCD (C) enantiomers into two peaks, while the meso forms d-(D) and ε-HBCD (E) only formed single peaks. Optical activities of different enantiomers have been assigned before . Fig. S4 displays LC-MS results from a C18-reverse phase column, on which HBCD diastereomers but not enantiomers can be resolved. Fig. S3 also indicates that substantial amounts of (À)b-HBCD (B) and ε-HBCD (E) are converted by LinA2. Accordingly, two PBCDen products (highlighted) were formed, eluting after 25.8 and 27.3 min from the b-cyclodextrin column. In addition, smaller amounts of (þ)b-HBCD were transformed to a PBCDen eluting after 29.1 min. Fig. S2 displays all 16 E-PBCDen configurations possible. Accordingly, 16 Z-PBCDens may be expected. Chromatographic retention N.V. Heeb, J. Hubeli, T. Fleischmann et al. Chemosphere 267 (2021) 129217 times of HBCDs and PBCDens on RP-and CP-columns are given in Table S1 (SM). We conclude that three of eight exposed HBCD stereoisomers fit into the active site of LinA2 and were converted to only three PBCDen stereoisomers. Thus, LinA2 was both, substrateand product-selective. In Fig. 3A and B, we delineate the stereochemical preferences of LinA1 and LinA2, which convert (þ)aand (À)a-HCH to the respective pentachlorocyclohexenes. In Fig. 3C, the LinA2-catalyzed transformation of (À)b-HBCD to 1E,5R,6R,9S,10R-PBCDen (3a) but not to its 1E,5S,6S,9R,10S-enantiomer (3b) is delineated. Fig. 3D displays the symmetrical ε-HBCD configuration with three mirror planes (dashed lines) and a C 3v -symmetry. Due to the three-fold symmetry axis, stereocenters at C1, C5 and C9 in R-configuration and those at C2, C6 and C10 in S-configuration are superimposable. As a consequence, only two HBr elimination products with the 1E,5R,6S,9R,10S (8a) and the 1E,5S,6R,9S,10R (8b) configuration can be expected. These PBCDens are enantiomers and, in principle, can be separated on a chiral-phase column (Fig. S3) but not on a RP column (Fig. S4). Chromatographic evidence indicates that only one PBCDen product was formed during the LinA2-catalyzed transformation of ε-HBCD (Fig. S3), while two products were obtained after LinA1 exposure (Fig. 4). Thus, the ε-HBCD transformation with LinA2 is stereo-and regio-selective too, like the ones of (À)a-HCH (Suar et al., 2005) and (À)b-HBCD (Heeb et al., 2015).

Stereoselectivity of the ε-HBCD transformation with three LinA variants
We used ε-HBCD to probe the stereoselectivity of the dehydrohalogenases LinA1, LinA2 and a triple-mutant LinATM with mutations L96C, F113Y and T133M. Other LinA variants, like those single-, double-and triple-mutants tested for a-HCHs by Shrivastava et al. (2017), might be interesting candidates too to identify those amino acids affecting the transformation of HBCDs.
Herein, we compared the kinetics and selectivity of these three LinA variants and exposed them to ε-HBCD at five different substrate/enzyme ratios [S o ]/[E o ]. Fig. 4 displays chromatographic evidence (CP) that all LinA variants converted some ε-HBCD to PBCDens. LinA2 converted nearly all of the given ε-HBCD in 32 h (Fig. 4A), while substantial amounts of substrate remained in case of LinA1 and LinATM ( Fig. 4B and C). Accordingly, large amounts of a single PBCDen product were formed by LinA2, while only small amounts of two PBCDen products were observed after LinA1 and LinATM exposure. These findings are confirmed from RP-LC-MS data (Fig. S5, SM) with the difference that both PBCDen enantiomers 8a and 8b appear as one single peak on the RP column. Small amounts of two PBCDen isomers can be detected in the starting material as well (Fig. 4, t ¼ 0 h). They eluted after 26.7 and 27.2 min from the CP column and were assigned to PBCDen 8a and 8b too. Initial rates were deduced from linear regression models shown in Fig. S6 (SM). As shown in Fig. 5, hyperbolic dependences are obtained when plotting initial rates v o (mmol L À1 h À1 ) versus initial substrate concentrations [S] o (mmol L À1 ), as expected for Michaelis-Menten kinetics. Observed (symbols) and modeled (lines) data of the ε-HBCD conversion with LinA2, LinA1 and LinATM are Fig. 3. Regio-and stereoselective dehydrohalogenations of (þ)a-HCH (A) by LinA1 and of (À)a-HCH (B), (À)b-HBCD (C) and ε-HBCD (D) by LinA2. The dehydrobromination of highly symmetrical ε-HBCD (D), with three mirror planes (dashed lines) and a C 3 -rotation axis, can lead to only two PBCDens. Elimination of bromines at C1, C5 and C9, which all are in Rconfiguration, lead to PBCDen 8b (D, right). Accordingly, bromine losses at C2, C6 and C10, which are in S-configuration, lead to the PBCDen enantiomer 8a (D, left). The LinA2catalyzed transformation of (À)b-HBCD (7b) to 1E,5R,6R,9S,10R-PBCDen 3a (C) and the one of (À)a-HCH to 1E,3R,4R,5S,6R-pentachlorocyclohexene (B) are stereoselective too. The (þ)-a-HCH enantiomer is preferentially transformed by LinA1 (A), but not by LinA2. Assigned HBCD and PBCDen structures refer to the numbering in Figures S1 and S2 (SM). Leaving halides are highlighted in magenta.  compared. Conversion of ε-HBCD was fastest with LinA2. Table S2 (SM) lists Michaelis-Menten parameters, deduced from these hyperbolic regression models (Fig. 5). Maximal reaction rates (v max ) of 1.11 ± 0.09, 0.39 ± 0.04 and 0.55 ± 0.09 mmol L À1 h À1 were not reached in the observed concentration range. Michaelis-Menten constants K M of LinA2, LinA1 and LinATM were 4.1 ± 0.5, 2.9 ± 0.4 and 5.3 ± 1.1 mmol L À1 , respectively (Table S2). Thus, K M of LinA2 and LinATM were similar, the one of LinA1 was about two times lower, indicating that LinA1 binds ε-HBCD tighter. Respective first-order rate constants k cat were 1.92 ± 0.16, 0.68 ± 0.07 and 0.95 ± 0.16 h À1 . Thus, ε-HBCD was converted three times faster by LinA2 than by LinA1, and conversion by the LinATM was half as fast as the one of LinA2.
Respective enzyme specificities k cat /K M decreased from 130 ± 25 to 66 ± 17 and 50 ± 19 L mol À1 s À1 for LinA2, LinA1 and LinATM, respectively. In other words, LinA2 had the highest specificity (Fig. 5). Table S2 also includes v max , K M and k cat /K M of the (À)b-HBCD conversion by LinA2, which were 0.17 ± 0.01 mmol L À1 h À1 , 0.47 ± 0.07 mmol L À1 and 178 ± 36 L mol À1 s À1 , respectively (Heeb et al., 2014). Thus, v max of the (À)b-HBCD conversion was about seven times slower than the one of ε-HBCD. Because K M was nine times lower, the overall enzyme specificity k cat /K M for (À)b-HBCD was slightly higher than the one for ε-HBCD.
We conclude that the three point mutations introduced in the substrate-binding site lowered the rate and product selectivity of LinATM to values observed for LinA1. In other words, LinA2 has the highest rate and product selectivity for ε-HBCD, while the triple mutant with lower selectivity, might preferentially bind larger substrates.

Regio-and stereoselective transformation of ε-HBCD by LinA2
Configurations of ε-HBCD (10) and both b-HBCD enantiomers (7a/b) are very similar (Fig. S1, SM). They differ only in one chiral center as indicated in Fig. 6 (blue circles). Inversion at C2 of (þ)b-HBCD (Fig. 6A) from S-to R-configuration results in ε-HBCD in the conformation given in Fig. 6C. Accordingly, an inversion at C2 of (À) b-HBCD (Fig. 6B) from R to S-configuration results in ε-HBCD in the enantiomeric conformation (Fig. 6D). Both ε-HBCD conformations shown in Fig. 6C and D are enantiomeric (dashed lines) and in equilibrium. In other words, both conformations are energetically equal and found in a 1/1 mixture. Thus, εand b-HBCDs have related configurations and conformations and may interacted likewise with LinA enzymes.
In analogy, we assume that both ε-HBCD conformations ( Fig. 6C and D) can bind to LinA2 like both b-HBCDs ( Fig. 6A and B). Thus, the enzyme catalyzes the cleavage of those bromine atoms highlighted in magenta (Fig. 6). Interestingly, both pathways starting with the ε-HBCD conformations shown in Fig. 6C and D lead to the same PBCDen product 8b with the 1E,5S,6R,9S,10R configuration (Fig. 6E). Thus, only one product is expected in silico for the LinA2catalyzed transformation of ε-HBCD, which is in accordance with experimental data (Fig. 4). Fig. 7A,B displays the activated (*) conformations of (þ)band (À)b-HBCD as they bind to LinA2 according to docking experiments (Heeb et al., 2015). These substrate-enzyme complexes seem reasonable. In case of the (À)b-HBCD-complex, the predicted PBCDen product 3a (Fig. S2) was in accordance with the observed product. The findings confirm that LinA2-catalyzed transformations of both b-HBCD enantiomers are regio-and stereoselective. Only those bromine atoms highlighted at positions C10 in (þ)b-HBCD (Fig. 6A) and at C5 in (À)b-HBCD (Fig. 6B), are recognized by the halide binding site. Both carbon atoms are in R-configurations and part of vicinal dibromides with left-handed synclinal orientations (-sc).
In analogy, we postulate that both ε-HBCD conformations ( Fig. 6C and D) also bind to LinA2 like (þ)band (À)b-HBCD. Respective enzyme-substrate complexes are displayed in Fig. 7C and D. Those bromides released by LinA2 bind to the halide-binding site (Fig. 7, red). As mentioned, both enzyme-substrate complexes ( Fig. 7C and D) lead to the same PBCDen product 8b with the 1E,5S,6R,9S,10R-configuration (Fig. 6E). Those bromine atoms of ε-HBCD differing from the ones of b-HBCDs are highlighted in magenta ( Fig. 7C and D). They are not in vicinity to the catalytic dyad and the halide-binding site and are not directly involved in the catalytic cycle but may have a role in substrate recognition.

Conclusions
The large-scale application of HCHs as insecticides, until the ban in 2009, has led to a global contamination and a chronic exposure of biota to low levels of these compounds. Prolonged exposure of certain Sphingobium bacteria to high HCH concentrations directed bacterial evolution and resulted in the appearance of new dehydrohalogenases like LinA1 and LinA2. LinA1 and LinA2 sequences are highly conserved, but substrate recognition differs. Shrivastava et al. (2017) studied the substrate specificity of several LinA mutants, among them single-, double-and triple-mutants, with respect to a-HCHs. Based on this study and our own work, we postulated that amino acids L96, F113 and T133 of LinA2 might affect substrate orientation of HBCDs too. With three point mutations (L96C, F113Y, T133M), we imbedded the substrate binding site of LinA1 into LinA2. These three mutations altered the kinetics (k cat ), the catalytic specificity (k cat /K M ) and the product selectivity.
The release of HBr from (À)b-, (þ)b-and ε-HBCD by LinA2 is both, regio-and stereo-selective leading to only one PBCDen product for each HBCD stereoisomer but not to the respective enantiomers. So far, (À)b-, (þ)b-and ε-HBCDs are the only brominated substrates known that are converted by LinA enzymes. Lately, we showed that also certain chlorinated paraffins are substrates for LinA2 (Heeb et al., 2019).
As shown, we could alter the substrate-and product-specificity of LinA enzymes with three point mutations. The exchange of leucine to cysteine (L96C) induced a volume reduction of À52 Å 3 in the substrate-binding site. Slightly more steric strain is introduced by F113Y (þ6 Å 3 ) and T133 M (þ31 Å3 ) mutations. Overall, more space is offered in the substrate binding site of the LinATM, which lowered both, the rate and selectivity of the dehydrohalogenation reaction. Other properties such as the polarity, the hydrogenbonding properties or the electronic structure might have changed too by these mutations. Specific mutations of these three positions seem promising for future studies and may help to design new dehydrohalogenases to convert other persistent organic pollutants. Thus, more extensive exploration of the role of amino acids near the active site is a prerequisite to further engineer LinA enzymes with changed substrate specificities. Well-designed dehydrohalogenases might be useful in bioremediation of contaminated sites (Lal et al., 2010).
Configurations of band ε-HBCDs are identical for 5 out of 6 stereocenters. Inversion of configurations at C2 (blue) of (þ) and (À)b-HBCDs from S to R and from R to S, respectively, results in those two ε-HBCD conformations shown in C and D. Assuming that these ε-HBCD conformations bind to LinA2 as well, only one transformation product is expected. PBCDen 8b (E) with the 1E,5S,6R,9S,10R configuration, but not its enantiomer PBCDen 8a (E) is the expected product. Vorkamp et al., 2012) may be explained by selective enzymatic transformations. This has been observed in-vitro for LinA and LinB (Heeb et al., 2013(Heeb et al., , 2015(Heeb et al., , 2018. In conclusion, despite the proven persistence of polyhalogenated pollutants at least some of these compounds can be dehalogenated by certain enzymes, expressed in bacteria of the Sphingobium family. By directed or random biological evolution, these enzymes may be further optimized to convert other POPs, regulated under the Stockholm Convention.  and two with synclinal (sc) orientations are present in (þ)b-(A) and (À)b-HBCD (B). Inversion at C2 of (þ)b-HBCD (A) from S to R produces ε-HBCD in the conformation C, with two dibromides in þsc and one in esc orientation. Accordingly, inversion at C2 of (À)b-HBCD (B) from R to S produces ε-HBCD in conformation D, with two dibromides in esc and one in þsc orientation. In each case (AeD), dibromides in esc orientations are bound to the active site (magenta) and leaving bromides in R-configurations are in close contact to the positively charged halide-binding site. The leaving bromides are coordinated by three nitrogen atoms in distances of 2.7e4.8 Å (dashed lines) of K20, W42 and R129. The catalytic dyad of D25 and H73 (magenta) acts as a base, binding hydrogen atoms from the backside, in a-positions to the leaving bromides. These four enzyme-substrate complexes (AeD) are similar with respect to active site geometry. They allow 1,2-diaxial eliminations of HBr leading to PBCDens 5b and 3a, when starting with (þ)band (À)b-HBCD (A,B). PBCDen 8b is the expected product for both ε-HBCD-enzyme complexes (C,D). manuscript was written by J.H. and N.V.H. All authors have discussed the results, reviewed and approved the manuscript.

Declaration of competing interest
The authors do not have any conflicts of interest with other entities or researchers regarding a publication of their data. Any opinions, findings, conclusions and recommendations expressed in this publication are those of the authors and do not necessarily reflect the view of the funding agencies.