A Detailed View on Geosmin Biosynthesis

The bacterial geosmin synthase is a fascinating bifunctional enzyme that has been discovered almost two decades ago. Several aspects of the cyclisation mechanism from FPP to geosmin are known, but a detailed picture of the stereochemical course of this reaction is unknown. This article reports on a deep investigation of the mechanism of geosmin synthase through isotopic labelling experiments. Furthermore, the effects of divalent cations on geosmin synthase catalysis were investigated. The addition of cyclodextrin to enzymatic reactions, a molecule that can capture terpenes, suggests that the biosynthetic intermediate (1(10)E,5E)‐germacradien‐11‐ol produced by the N‐terminal domain is passed to the C‐terminal domain not through a tunnel, but rather through release into the medium and uptake by the C‐terminal domain.


Introduction
Geosmin (1, Figure 1) is an earthy odorant that is responsible for the smell of freshly ploughed earth. [1] The compound is produced by soil bacteria and was first isolated from Streptomyces griseus by Gerber and Lechevalier in 1965, [2] followed by its structure elucidation shortly after. [3] Today the compound is known to be very widespread and was reported from actinobacteria, [4] cyanobacteria, [5] myxobacteria, [6] ascomycete fungi, [7] basidiomycota, [8] and amoebae. [9] Geosmin has also been reported from plants including the liverwort Symphyogyna brongniartii, [10] mosses, [11] buckwheat (Fagopyrum tataricum), [12] Zea mays, [13] and beetroot where it adds to its typical earthy flavour. [14] In rainbow trout, [15] shrimps, [16] and molluscs [17] it may be a contaminant of microbial origin which is a particular problem in aquaculture. Geosmin is also a constituent of the defence secretions of the polydesmid millipede Niponia nodulosa. [18] Several derivatives of geosmin have been reported from natural sources, including dehydrogeosmin (2) from the flower scent of the cactus Ributia marsoneri, [19] the epoxide 3 from the liverwort Lophocolea bidentata, [20] and oxygenated compounds such as 4 and 5 from an endophytic streptomycete. [21] The biosynthesis of geosmin was a long standing problem. Initial speculations suggested that geosmin may be a degraded eudesmane sesquiterpene. [3] Subsequent feeding experiments with radioactively labelled precursors to Streptomyces antibioticus showed incorporations from [1-14 C] and [2][3][4][5][6][7][8][9][10][11][12][13][14] C]acetate, but not from [methyl-14 C]methionine. [22] Based on the identification of the cometabolites (1(10)E,5E)-germacradien-11-ol (6) and dihydroagarofuran (7) from Streptomyces citreus a first detailed biosynthetic pathway through 6 and 7 as intermediates was proposed, but no satisfying mechanistic explanation was given. [23] At the time of the discovery of the geosmin synthase and its coding gene in 2003, [24] mechanistic proposals were raised that include oxidative degradation. [24b,25] Finally, based on feeding experiments with ( 2 H 10 )leucine, (4,4,4,5,5,5-2 H 6 )dimethylacrylate, and (4,4,6,6,6-2 H 5 )mevalonic acid lactone a mechanism was developed that proceeds through the cyclisation of FPP to 6, followed by reprotonation induced cyclisation to C that can undergo a retro-Prins reaction to the octalin 9 and acetone (Scheme 1); reprotonation to D, a 1,2-hydride shift to E and capture with water lead to 1. [26] A deep investigation of the geosmin synthase from Streptomyces coelicolor (ScGS), an enzyme exhibiting two functional domains, through sitedirected mutagenesis and in vitro experiments revealed that the N-terminal domain converts FPP into 6, while the C-terminal domain further converts this intermediate into 1. [27] Isotopic labelling experiments revealed that the deprotonation of A to 8 proceeds with specific loss of the 1-pro-S hydrogen of FPP. This is the same hydrogen that undergoes a 1,3-hydride shift from A to F on the pathway to the side product germacrene D (10). [28] This mechanism was further supported by the rigorous structure elucidation of 9 through total synthesis [29] and the capture of acetone from an in vitro conversion of FPP. [30] Today also the crystal structure of the N-terminal domain of ScGS is known, [31] but full structural insights into the bifunctional enzyme have still not been obtained. Here we report on the functional characterisation of the geosmin synthase from Allokutzneria albata (AaGS) and investigations on the cyclisation mechanism by isotopic labelling experiments.
The cyclisation mechanism to 1, its intermediate 6 and the side product 10 was studied in detail through isotopic labelling experiments (Table S5), with the aim to unravel the precise stereochemical course for every elementary step along the cyclisation cascade. The enzymatic conversion of (R)-and (S)-(1-2 H)FPP [36] with AaGS and GC/MS analysis of the products revealed a specific loss of the 1-pro-S hydrogen in 6 and 1 and a specific migration of the same hydrogen into the iPr group of 10 ( Figure S27), confirming earlier results with ScGS. [28] Furthermore, the usage of (R)-and (S)-(1-13 C,1-2 H)FPP [37] established that the 1-pro-R hydrogen remains bound to the original carbon in 1 and 6, as indicated by a slightly upfield shifted triplet for C-6 resulting from 13 C-2 H spin coupling ( Figure S28). Moreover, HSQC analysis of the product mixture obtained from (R)-(1-13 C,1-2 H)FPP demonstrated the α orientation of the deuterium atom in 1 ( Figure S29). This is in line with a formation of 1 from 9 by reprotonation and a suprafacial 1,2-hydride shift, disfavouring a hypothetical role of 9 as a shunt product rather than intermediate. Interestingly, the 13 C NMR spectra for the enzymatic conversions of (12-13 C)FPP and (13-13 C)FPP (prepared   with (9-13 C)GPP, IPP and FPPS) with AaGS indicated that there was a minor exchange of 13 C labellings between C-12 and C-13 in the biosynthesis towards 6, while this phenomenon was not observed in the biosynthesis of the side product 10 which shares the same cationic intermediate A ( Figure S30). This finding may be explained by a protonation induced opening of the three-membered ring in 8 to produce a cation at C-11 in B that is preferentially attacked from one side, but to a minor extent also from the other side. An alternative rotation of the iPr group in A is less likely, because this should also lead to an exchange of labelling between C-12 and C-13 in 10. The 1,2hydride shift from D to E was also directly investigated using geranyl diphosphate (GPP) that was coupled with (1-13 C,2,2-2 H 2 )IPP by S. coelicolor FPP synthase [38] to yield (1-13 C,2-2 H)FPP. Its further conversion into labelled 1 resulted in a triplet signal for C-6, confirming the migration of deuterium to this labelled carbon ( Figure S31). The biosynthesis of 1 proceeds with a total number of three reprotonation steps: Intermediate 8 is reprotonated at C-4, 6 becomes reprotonated at C-1, and reprotonation of 9 happens at C-7. These reactions can be studied using a 13 C-label in the substrate FPP at the reprotonated carbon in conjunction with incubation in an deuterium oxide buffer. Moreover, HSQC analysis of the labelled product 1 can show the stereochemical course of the reprotonation step. Enzymatic conversions of (3-13 C)FPP, (6-13 C)FPP and (10-13 C)FPP [39] with AaGS in deuterium oxide confirmed all three reprotonation steps and established deuterium incorporation into H-4, H-1α and H-7β ( Figures S32-S34).
Dionigi reported that the production of 1 in Streptomyces albidoflavus is increased in the presence of copper sulfate. [40] On the contrary, Schrader and Blevins found depleted levels of 1 in Streptomyces halstedii grown in the presence of divalent zinc, iron or copper. [41] These effects could be indirect or could be direct results of the presence of these divalent ions on the activity of geosmin synthase. To investigate the effect of different divalent cations, AaGS was incubated with FPP without any metal ions and with MgCl 2 , CaCl 2 , FeCl 2 , NiCl 2 , ZnCl 2 , MnCl 2 and CuCl 2 . These experiments revealed a strict dependency on Mg 2 + that can only be substituted with Mn 2 + , resulting in a relative activity of 17.1 � 3.3 % in comparison to Mg 2 + (set to 100 % for the sum of the four main products 1, 6, 9, and 10 of AaGS, Table S6). No activity was observed with any of the other tested salts or without addition of a divalent cation.
The effects of combinations of MgCl 2 (5 mm) and additional divalent cations (1 mm) on the relative production of compounds 1, 6, 9 and 10 were also investigated ( Figure 4 and Table S7, the sum of all four compounds was set to 100 % for the incubation with MgCl 2 only). To visualise specific changes for individual compounds their relative production is given in Figure S35. Addition of CaCl 2 resulted in a slightly enhanced production of all compounds (107.5 � 3.3 %) and especially of 6 (120.8 � 3.7 %), suggesting that the activity of the N-terminal domain is more strongly increased than the activity of the Cterminal domain, which results in the accumulation of 6. In contrast, the addition of FeCl 2 gave an overall reduced activity (74.1 � 3.6 %), with a particularly strong decrease of 10 (40.4 � 3.0 %) and only a minor effect for 6 (88.4 � 4.8 %). The presence of Fe 2 + may shift the ratio for the reaction paths of the Nterminal domain towards the main product 6. Administration of CoCl 2 caused a slightly reduced activity (81.6 � 1.6 %) with similar effects on all four compounds. Similar observations were made with NiCl 2 (82.1 � 6.7 %) and ZnCl 2 (81.2 � 8.8 %). The relative production with MnCl 2 dropped to 46.8 � 4.2 %. In first instance, this result seems unexpected, because this is the strongest decrease with any of the metal cations tested, although Mn 2 + was the only divalent cation that could substitute for Mg 2 + . An explanation for this finding may be that Mn 2 + is besides Mg 2 + the only divalent cation that can bind to the active site of AaGS, and the competing bindings between Mn 2 + and Mg 2 + may lead to a reduced activity. Finally, with CuCl 2 a moderately reduced activity was found (80.3 � 6.0 %). In particular, the production of 1 (47.8 � 4.8 %) and 9 (50.8 � 3.1 %) was suppressed, while the amounts of 6 were increased (111.4 � 7.4 %), suggesting that Cu 2 + specifically inhibits the Cterminal domain which consequently results in the accumulation of 6.
A special feature of bifunctional enzymes or enzyme complexes are substrate tunnels that can be used to transport the product from the active site of one domain or subunit to the active site of the other domain or subunit. Substrate tunnels are e. g. known for the transport of indole between the α and β subunits of tryptophan synthase, [42] or of ammonia in bifunctional enzymes in which one domain releases ammonia through hydrolysis of the amide function of Gln and another domain incorporates ammonia into a substrate molecule. Examples include glutamate synthase [43] or formylglycinamide ribonucleotide amidotransferase involved in the biosynthesis of purine nucleotides. [44] In contrast, cryo-EM studies on the bifunctional fusicoccadiene synthase containing a prenyltransferase domain for the biosynthesis of geranylgeranyl diphosphate (GGPP) and a terpene synthase domain for its cyclisation to fusicoccadiene revealed that in this enzyme substrate channeling does not proceed through a tunnel. Instead, substrate channeling is enabled by the proximity of multiple terpene synthase domains that surround an oligomeric core of prenyltransferase domains. [45] A similar situation was recently reported for farnesylfarnesyl diphosphate (FFPP) dependent bifunctional fungal triterpene synthases. [46] To investigate whether a substrate tunnel between the N-and the C-terminal domains of AaGS may be involved in channelling of 6 from one to the other active site, AaGS and FPP were incubated with or without the addition of β-cyclodextrin (β-CD). Terpenes can form host-guest complexes with β-CD, [47] and if 6 would be released into the medium by the N-terminal domain of AaGS, the addition of β-CD to the enzyme reaction could lead to a partial capturing of 6 that should consequently result in a lower production of 1. This turned out to be the case: While the production of 6 and 10 upon addition of β-CD was similarly high as without β-CD, only traces of 1 and 9 were detected in the presence of β-CD ( Figure 5).

Conclusion
In summary, we have functionally characterised a new geosmin synthase from Allokutzneria albata (AaGS), and resolved the biosynthesis of geosmin (1), its intermediate (1(10)E,5E)-germacradien-11-ol (6) and its major side products 9 and 10 in all detail through isotopic labelling experiments. The results showed the sites and the stereochemical course of three reprotonation steps and the fate of the geminal methyl groups of FPP that are slightly exchanged. Moreover, direct evidence for a 1,3-hydride migration and a late stage 1,2-hydride shift were obtained. The catalytic activity of AaGS was also shown to strictly depend on Mg 2 + or Mn 2 + , while other divalent cations cannot substitute for these metal ions. Reports in the literature mentioned an increased or decreased production of 1 during growth of geosmin producers in the presence of Cu 2 + . These effects can have several reasons, but we showed here that Cu 2 + has indeed a direct effect on AaGS and leads to decreased yields of 1. Incubations of the bifunctional AaGS in the presence of β-CD, a molecule that can bind terpenes, resulted in a disrupted production of 1, while 6 was not affected. The intermediate 6 is the product of the N-terminal domain that is passed on to the C-terminal domain for further conversion into 1. The effect of β-CD suggests that 6 is not transferred from the N-terminal to the C-terminal domain through a tunnel, but rather through the medium where is becomes captured by the added β-CD. However, details of the interaction of the two domains of geosmin synthase remain unknown, also because still no structural data for the full bifunctional enzyme are available. This leaves room for additional future studies on this unique bacterial terpene synthase.

Experimental Section
Phylogenetic tree construction: The phylogenetic tree of geosmin synthase homologs was constructed from 1062 amino acid sequences that were identified by BLAST search using the amino acid sequence of AaGS as a probe (accession number WP_ 011030632). The tree was constructed using the tree builder function of Geneious (alignment type: global alignment with free end gaps, cost matrix: Blosum45, genetic distance model: Jukes-Cantor, tree build method: neighbor-joining, gap open panelty: 8, gap extension penalty: 2).

Gene cloning and plasmid construction:
The desired gene was obtained from freshly isolated genomic DNA from Allokutzneria albata by PCR using Q5-High-Fidelity DNA polymerase (New England Biolabs, Ipswich, MA, USA) and the primers WP030430635_ Fw and WP030430635_Rv (Table S1). PCR standard conditions were used (initial denaturation at 98°C for 1 min, 35 cycles with denaturation at 98°C for 15 sec, annealing at 62°C for 30 sec and elongation at 72°C for 68 seconds, final elongation step at 72°C for 2 min). The PCR products together with the linearised pYE-Express shuttle vector were used for a yeast homologous recombination by a standard protocol using PEG, LiOAc and salmon sperm DNA. [48,49]  The transformed Saccharomyces cerevisiae cultures were plated on SM-URA plates and grown for 3 days at 28°C. Colonies were collected from the plates and plasmid DNA was isolated using Zymoprep Yeast Plasmid Miniprep II (Zymo Research, Irvine, USA). The isolated plasmid DNA was used for electroporation of E. coli BL21 (DE3) electrocompentent cells. The transformed E. coli was grown overnight at 37°C on LB agar plates containing kanamycin (50 μg mL À 1 ). Single colonies were selected to inoculate LB medium with kanamycin (8 mL). The resulting cultures were grown for 24 h to isolate plasmid DNA, yielding plasmid pYE_WP030430635 which was checked by analytical digest and by sequencing.
Gene expression and enzyme purification: For gene expression a preculture of the E. coli transformants harbouring the plasmid pYE_ WP030430635 was grown overnight at 37°C in LB medium containing kanamycin. The expression cultures were then inoculated using 20 mL L À 1 of preculture, followed by culturing at 37°C with shaking until an OD 600 = 0.4-0.6 was reached. The cultures were cooled to 18°C and protein expression was induced by addition of IPTG (400 mm in water, 1 mL L À 1 , final concentration 0.4 mm). The expression cultures were shaken overnight at 18°C and then centrifuged at 3.600 × g for 40 min. The supernatant was discarded and the cell pellet was resuspended in binding buffer (10 mL L À 1 culture; 20 mm Na 2 HPO 4 , 500 mm NaCl, 20 mm imidazole, 1 mm MgCl 2 , pH = 7.4, 4°C). The resulting suspension was subjected to ultra-sonication for cell lysis. The cell debris was removed by centrifugation (14.610 × g, 15 min) and the supernatant was loaded onto a Ni 2 + -NTA affinity chromatography column (Super Ni-NTA, Generon, Slough, UK). The column was washed with binding buffer (2 × 10 mL L À 1 culture), and the desired His-tagged protein was eluted using elution buffer (10 mL L À 1 culture, 20 mm Na 2 HPO 4 , 500 mm NaCl, 500 mm imidazole, 1 mm MgCl 2 , pH = 7.4, 4°C). Fractions containing protein were pooled, analysed by Bradford assay to determine the protein concentration (0.6 mg mL À 1 ) [50] and by SDS-PAGE ( Figure S1), and used for analytical-scale incubation experiments.