*5.2.* α*-Ketoglutarate-Dependent Halogenases*

Table 4 shows different natural products that are formed by the iron(II)-α-ketoglutarate-dependent (Fe/αKG)-halogenases. Despite the huge variety in the product structures they share one common feature, which is the halogen at a sp3-carbon centre. Hence, the Fe/αKG-halogenase is not limited to nucleophilic substrates like the previous described enzymes. They belong to the Fe/αKG-dependent oxygenase superfamily. The superfamily is known for different transformations such as hydroxylation [138], halogenation [139], desaturation [140], or can be used for the production of ethylene [141]. They all share a structurally conserved metal-binding motif, which in the case of the halogenase developed an active centre that is eventually able to bind a haloge n [139]. The proposed catalytic mechanism of Fe(II)/α-KG-dependent-halogenase is illustrated in Figure 16.


**Table 4.** Examples of heterologously expressed Fe(II)/αKG-dependent halogenases.

Based on the proposed radical C-H functionalization two classes of enzymes have so far been identified. The first such reported Fe/αKG-dependent halogenase is the tailoring domain SyrB2 of the multimodular nonribosomal peptidsynthetase (NRPS) from *Pseudomonas syringae* pv. *syringae* B301D [133,144]. These NRPS-associated halogenases produce a diversity of secondary metabolites such as the chlorinated biosurfactant syringomycin E (**31**), which is characterized by a selective monochlorinated threonine in its structure [128,139]. Another example is the highly selective diand trichlorination of solely one of the diastereotopic methyl groups of leucine by a combination of BarB1 and BarB2, which serves as a precursor for the natural compound barbamid (**32**) in the marine cyanobacteria *Lyngbya majuscula* [142,153].

**Figure 16.** Proposed mechanism for halogenation reaction by Fe(II)/αKG-dependent halogenase via a radical C-H functionalization [142]. The highly reactive Fe(IV)-oxo (haloferryl) intermediate is produced by decarboxylation of αKG to succinate through an oxygen attack. Subsequently abstraction of a hydrogen-atom from the substrate leads to an energetically favourable rearrangement towards Fe(III). Rebound reaction with chloride was shown to depend on the distance and orientation of the substrate [143]. The catalytically cycle is re-established by the hexa-coordinated Fe(II) with water molecules, chloride and histidine.

Recently Moosmann et al. identified different αKG-halogenase homologues and their natural products that are produced via a NRPS (non-ribosomal peptide synthetase) pathway. The halogenase were identified by screening the genomic sequence of the cyanobacterium *Fischerella* sp. PCC 9339 based on feature comparison. Using this approach, the authors were able to distinguish between a Fe/αKG oxygenase and a corresponding halogenase [143,154]. However, large NRPSs characteristically bind their substrates through an aminoacetylated peptidyl-carrier protein and have a narrow substrate scope [144,155]. Furthermore, they generally showed a low total turnover number, which may result from the well-known autoxidation of Fe(II) to Fe(III) and hence an auto-inactivation of the enzyme [145,156]. In case of SyrB2, total turnovers of 7 ± 2 were observed [133,144] This limits the possibility to modify the enzymes in order to use them as suitable biocatalysts for different unnatural

substrates. With the discovery of a new Fe(II)/αKG-dependent halogenase (WelO5) by Hillwig and Liu, it was possible to expand the class towards unbound substrates. WelO5 is capable of late stage halogenation in a regio- and stereoselective manner of different derived isoprenoid-indole alkaloids in the cyanobacterium *Hapalosiphon welwitschii* (see Table 4) [137,148]. WelO5 showed also a higher robustness and catalysis of approximately 75 turnovers in total [137,148]. Strategies such as adding the cosubstrates consecutively or adding antioxidants like catalases or DTT could increase the turnover number. The narrow substrate scope of WelO5 was tailored in order to have an increased substrate scope like the homolog AmbO5 [138,149]. Most modifications were at the external helix, which is responsible for closing the entry of the active site upon binding of the substrate. It can be assumed that the helix is partially involved in the substrate recognition and specificity [138,149]. A recent publication from Hayashi et al. showed a WelO5 variant with a reshaped active site that led to improved kinetics and an expanded substrate scope, which applies beyond the native indole alkaloid-type substrates [141,152]. This provides the possibility for targeted enzyme-engineering and a basis for further improvements in substrate scope. One possibility is the establishment of nitration and azidation as already shown for SyrB2 [157]. In this regard, it has been shown that WelO5 is able to incorporate the unnatural halide Br− [158]. One drawback of engineering Fe/αKG-dependent halogenases is the hydroxylation as a competitive side reaction [155]. Mitchell et al. used this approach backwards and modified a monooxygenase SadA towards a halogenase [159]. This serves as a proof of concept that with increasing understanding of the reaction mechanism and the involved amino acids the superfamily of monooxygenase can be used as a versatile toolbox in biotechnology. In the future, this may lead to the use of different variants of the very same enzyme for different transformations. Table 4 shows an overview of different characterized Fe/αKG halogenases and their main published features. Excluded are, for example, halogenase modules of NRPS, where the halogenation is necessary for the subsequently formation of cyclopropane such as in case of CurA [160] or CmaB (see Figure 17) [161].

**Figure 17.** Schematically example for formation of cyclopropane initiated by CmaB through halogenation.

## *5.3. Fluorinase*

In contrast to the other described enzymes, the diversity of natural products in case of the fluorinase stem from only one characterized enzyme class to date. The involved enzymes are *S*-adenosylmethionine (SAM) dependent. The first characterized representative was FlA (5 -fluoro-5 -deoxyadenosine synthase) from [162]. The overall family of this enzymes is also able to chlorinate or hydroxylate SAM, as described in detail elsewhere [163]. Within the catalytic cycle fluoride acts as a nucleophile in a SN2-reaction, where it attacks the 5 -carbon of SAM-ribose [164]. In order to act as a nucleophile, fluoride requires to lose its solvation shell. This is achieved in a two-step desolvation with a combination of electrostatic stabilization and hydrogen bonding. In the first step, fluoride is binding to the active site and exchanges water molecules of its shell in order to form hydrogen bonds with the enzyme. Upon binding of SAM the desolvation of fluoride is complete. The electropositive 5 -carbon attached to the sulfur group in SAM coordinates with the fluoride [165,166]. This electrostatic stabilization facilitates the nucleophilic attack of the fluoride and C−F-bond formation of the reactive 5 -fluoro-5 deoxyadenosine (**33**, 5 -FDA) intermediate [166]. Subsequently, 5 -FDA (**33**) is further metabolized in order to generate a variety of compounds as shown in Figure 18A [162]. However, this also represents a major obstacle for the application of these enzymes to unnatural small organic molecules, since the product formation follows a cascade of enzymatic steps. Eustáquio et al. tried to use this enzyme for the production of fluorosalinosporamide, an unnatural

analog of salinosporamide, which is fluorinated rather than chlorinated, however, the yield was moderate [167]. Nevertheless, different approaches have been implemented to increase the substrate scope and use the enzyme as rather flexible tool for medical applications. Besides the ability to fluoride compounds, fluorinase is also able to exchange a chloride at the 5 -carbon of the SAM ribose ring by a fluoride and form 5 -FDA (**33**, Figure 18B) [168]. This overall trans-halogenation reaction was used for late-stage fluorination for the production of radiolabeled imaging reagents. Recently, different pre-targeting strategies have been developed for treatment and imaging of different diseases. Those include e.g., radiolabeling of the human A2A adenosine receptor [169], a prostate cancer-related membrane protein [170] or the combined application of biotin and tetrazine-conjugate with antibodies [171]. In all cases, it was shown that the fluorinase (FlA) accepts substrates with different moiety at C-2 of the adenine ring. So far two crystal structures of fluorinase homologs from *S. cattleya* and *Streptomyces* sp. MA37 are known and they both have a high structurally conformity [172]. In general all five known fluorinases have a high similarity of over 80% and show similar kinetic profiles [173]. Through a directed evolution approach of FlA1, different crucial amino acids for substrate binding, halide binding and hence activity were identified [174]. Additionally, the variants were tested with different unnatural substrates [175]. It was shown that the tolerance for the *wild type* (*wt*) enzyme is limited to C-2 modified substrates. However, generated variants of FlA1 also demonstrated an activity for an unnatural substrate, which were modified at C-6 positions of the adenine ring with a chlorine group [175]. These findings show that despite a narrow substrate scope of the fluorinase, it was possible to successfully apply different unnatural substrates and lay a base for directed evolution as means to use small organic compounds as substrates. However, the dependence of electrophilic substrate structures remains a drawback for the nucleophilic attack of fluoride. The crystal structure with an unnatural substrate (containing difluoromethyl groups) confirmed the necessity of geometry for activating the fluorine atom for substitution [176]. This outlines the challenge to use fluorinases as a versatile tool to generate fluorinated pharmaceutical compounds. Nevertheless, by means of designing appropriate leaving groups in combination with enzyme engineering, fluorinases could be used as tool for future generation of fluorinated pharmaceutical compounds. Data of known fluorinases are displayed in Table 5.


**Table 5.** Examples for heterologously expressed fluorinase. Kinetic data representing the conversion of 5 -ClDA into *S*-adenosylmethionine (SAM).

**Figure 18.** (**A**) Schematically sequential mechanism for the F–C bond formation catalysed by the fluorinase from *S. cattleya* and some products of subsequently cascade reaction. Dashed lines represents hydrogen bonding contacts with amino acids in the active pocket or water. (**B**) Reaction scheme of fluorinase-mediated trans-halogenation. Rest R marks position of usual derivation.

#### **6. Conclusion on Halogens in Active Agent (Syntheses)**

As we have seen in the previous paragraphs halogens are very important to many active agents as a functional moiety *per se* due to their physico-chemical properties such as bulkiness, latent polarization and as important binding partners because of halogen bonds. Organic halogen compounds are, furthermore, instrumental for synthetic purposes in terms of being good leaving groups and facilitating cross-metatheses by halogen-metal-exchanges. Nevertheless, these indispensable advantages have to be bought at a high price; namely the energy-intensive production of very toxic and hazardous chlorine gas. The reduction in energy consumption must mainly be managed by technical improvements of the chloralkali-process and enzymes are likely not able to make a significant impact. A major reason is that the majority of chlorinated compounds are necessary for different types of plastic materials (e.g., PVC) and solvents [51]. Enzymes are limitedly applicable in those areas of bulk chemicals, but there is a potential for fine chemicals. Even though halogenating enzymes will not replace conventional chlorine production, it is worth taking a look at this group of enzymes or rather at these groups of enzymes, because nature has invented these amazing enzymes at least six times. The expectations of these biocatalysts are that the conversions become environmentally more benign, the processes skip hazardous compounds such as chlorine gas and that conversions get more selective. However, the research in the field of halogenating enzymes is still at the beginning. Consistent enzymologic data such as kinetic data, measurements on stability or even well studied mutant libraries are rarely available. Many halogenating enzymes from eukaryotic sources suffer from expression

challenges. Nevertheless, these enzymes open up a wide horizon of possibilities. Enormous genome data are revealing more and more halogenating enzymes and even new classes of halogenating enzymes cannot be excluded at present. Thus, there is a need for detailed and systematic research to employ halogenating enzymes for active agent synthesis, to alter their substrate scopes and enhance their process stability.

**Author Contributions:** T.C. conceived this review and relegated the manuscript to a complete work. A.V.F., J.G., and N.H. carried out the initial research and compiled the various chapters. All authors contributed equally to this review.

**Funding:** This research was funded by Deutsche Forschungsgemeinschaft grant number 369034981, BioSC(CombiCom) Ministry of Culture and Science North-Rhine-Wesfalia grant number 313/323-400-00213, and European Regional Development Fund: CLIB-Kompetenzzentrum Biotechnologie.

**Acknowledgments:** For their fruitful discussions we want to thank *Jörg Pietruszka*, *Nora Bitzenhofer*, *Esther Breuninger*, and *Denise Detlof* as well as the entire institute staff for their support. We want to thank our sponsors for the financial support. AF was funded by the *European Regional Development Fund (ERDF)* within the '*CLIB-Kompetenzzentrum Biotechnologie*'. The scientific activities (JG) of the *Bioeconomy Science Center* were financially supported by the *Ministry of Culture and Science* within the framework of the *NRW Strategieprojekt BioSC* (No. 313/323-400-00213). NH is funded by the *Deutsche Forschungsgemeinschaft* (369034981). Furthermore, we want to thank the *Forschungszentrum Jülich GmbH* and the *Heinrich Heine University Düsseldorf* for their generous support.

**Conflicts of Interest:** The authors declare no conflict of interest.
