Oligomerization engineering of the fluorinase enzyme leads to an active trimer that supports synthesis of fluorometabolites in vitro.

The fluorinase enzyme represents the only biological mechanism capable of forming stable C-F bonds characterized in nature thus far, offering a biotechnological route to the biosynthesis of value-added organofluorines. The fluorinase is known to operate in a hexameric form, but the consequence(s) of the oligomerization status on the enzyme activity and its catalytic properties remain largely unknown. In this work, this aspect was explored by rationally engineering trimeric fluorinase variants that retained the same catalytic rate as the wild-type enzyme. These results ruled out hexamerization as a requisite for the fluorination activity. The Michaelis constant (KM ) for S-adenosyl-l-methionine, one of the substrates of the fluorinase, increased by two orders of magnitude upon hexamer disruption. Such a shift in S-adenosyl-l-methionine affinity points to a long-range effect of hexamerization on substrate binding - likely decreasing substrate dissociation and release from the active site. A practical application of trimeric fluorinase is illustrated by establishing in vitro fluorometabolite synthesis in a bacterial cell-free system.

Introduction

Organohalides (i.e., organic molecules containing halogens) are essential chemical building blocks for the synthesis of pharmaceuticals, materials and agrochemicals, an occurrence highlighted by the fact that over 25% of blockbuster drugs display fluorine (F) atoms in their structures (Martinelli and Nikel, 2019; Mei et al., 2019; Johnson et al., 2020). F is the most electronegative element in the Periodic Table, and the C–F bond is the strongest covalent bond involving C (O'Hagan, 2008; Cros et al., 2022). Substituting hydrogen (H) atoms with F has a relatively low steric impact – however, F exchange greatly influences the acidity/basicity, hydrophobicity, conformation and reactivity of organofluorine compounds relative to the non‐fluorinated hydrocarbon structures (Purser et al., 2008; Pimviriyakul et al., 2020; Wu et al., 2020). These modifications translate into improved pharmacokinetic parameters such as metabolic stability, increased binding to target molecules and/or improved membrane permeability when compared to the analogous non‐fluorinated compound (Meanwell, 2018; Johnson et al., 2020). Organic chemistry methods exist for the synthesis of organofluorines and, more often than not, these require hazardous and noxious reagents. It follows that there are attractive prospects for the development of a biotechnology in this area (Cheng and Ma, 2021), but this goal is presently challenged by very limited access to appropriate enzymatic routes towards organofluorine biosynthesis. As such, developing a biotechnology for the site‐selective introduction of F into structurally diverse molecules would circumvent the harsh chemistry associated with chemical fluorination, and remains one of the great goals in metabolic engineering (Wu et al., 2020). So far, the fluorinase [FlA, 5'‐fluoro‐5'‐deoxyadenosine (5'‐FDA) synthase], isolated from Streptomyces and related Gram‐positive bacterial species, is the only known enzyme capable of forming stable C–F bonds (O'Hagan et al., 2002; Dong et al., 2004). FlA catalyses the formation of 5'‐FDA through a nucleophilic attack of the fluoride ion (F–) on the electrophilic C5' carbon of S‐adenosyl‐l‐methionine (SAM, the universal methyl donor), thereby releasing l‐methionine (Met) through an SN2 reaction mechanism (Fig. 1A). FlA also catalyses the formation of SAM from 5'‐chloro‐5'‐deoxyadenosine (5'‐ClDA) and Met as the substrates (O'Hagan and Deng, 2015; Sun et al., 2018) – that is, in the reverse reaction direction.

Fig. 1

A. Reactions catalysed by the fluorinase enzyme. Abbreviations are as follows: Cl–, chloride ion; F–, fluoride ion; SAM, S‐adenosyl‐l‐methionine; Met, l‐methionine; 5'‐ClDA, 5'‐chloro‐5'‐deoxyadenosine; and 5'‐FDA, 5'‐fluoro‐5'‐deoxyadenosine.

B. A 3D model of the FlA fluorinase enzyme from Streptomyces sp. strain MA37. The amino acid residues sitting at the trimer‐trimer interface, critical for the formation of the hexameric form, are highlighted in the inset.

FlA has been successfully implemented in vitro (Sun et al., 2016; Carvalho and Oliveira, 2017) for a number of practical applications, for example, in the preparation of 18F‐labelled peptides for positron emission tomography (Liang et al., 2013). Additionally, this halogenase has been used for in vivo fluorometabolite biosynthesis in engineered bacterial cell factories (Eustáquio et al., 2010; Calero et al., 2020; Markakis et al., 2020). Yet, the applications of FlA have been hampered by its narrow substrate scope and low catalytic efficiency (with a turnover number k cat = 0.08–0.26 min–1; Deng et al., 2004; Cobb et al., 2006; Nieto‐Domínguez and Nikel, 2020). Crystal structures of FlA indicate that the enzyme forms a hexamer consisting of a dimer of trimers (Fig. 1B; Dong et al., 2004; Deng et al., 2006). The FlA fluorinase has also been described to exist as a hexamer in its native form in solution (Deng et al., 2014). According to this interpretation, three active sites are situated at the interfaces between adjacent monomers in the trimer, while the trimer‐trimer interface has no obvious catalytic function (Fig. 1B). This observation on the structural features of FlA, together with the fact that a homologous chlorinase enzyme (SalL) forms a functional trimer while displaying a k cat/K M specificity constant 100‐fold higher than that of the fluorinase (Eustáquio et al., 2008), led us to investigate the role of FlA hexamerization on catalysis. The few other fluorinase enzymes described in the literature have not been characterized in detail thus far (e.g., by crystallographic analysis), and the oligomerization status of these halogenases remain to be explored (Cheng and Ma, 2021). On this background, in this article we report our efforts towards rational deconstruction of the hexameric form of the FlA fluorinase. By introducing a number of rationally engineered mutations in the enzyme structure, we show how the formation of the hexamer is not a requisite for the catalytic activity. Moreover, we demonstrate that trimeric variants of FlA can be harnessed for the biosynthesis of fluoronucleotides in an in vitro, cell‐free bacterial system as a practical application example.

Results

Rational design, construction and testing of FlA mutants

With a view to disrupt hexamerization of the fluorinase while preserving intact trimers, we designed eight FlA variants guided by sequence alignments and the crystal structures of the archetypal FlA from Streptomyces sp. strain MA37 [PDB ID 5LMZ] (Deng et al., 2014) and the SalL chlorinase from Salinispora tropica [PDB ID 2Q6K] (Eustáquio et al., 2008). All of the engineered mutations are located at the trimer‐trimer interface of the fluorinase, and they were designed to substitute FlA residues with the corresponding SalL amino acids at key positions while considering electrostatic forces as part of the design principles (Table 1). We constructed these mutant versions of FlA by mutagenesis PCR (see Experimental Procedures), introducing either point mutations (individually or in combination) or by insertional mutagenesis. Four of these engineered variants (which were termed FlA‐K237E, FlA‐D241A, FlA‐K237E/N287P and FlA‐242_243InsLS) could be successfully produced in Escherichia coli BL21 and purified with high protein mass yields. We started the characterization of the enzyme variants by testing the ability of each FlA mutant to mediate the forward (i.e., fluorination) and reverse (i.e., SAM formation) reaction by exposing the purified enzymes to the corresponding substrates. In these experiments, either SAM or 5'‐ClDA were used as the substrate to assess the forward or reverse reaction catalysed by FlA respectively. The reaction mixture also contained KF (as the source of F– ions for the forward reaction), l‐Met (for the reverse reaction starting with 5'‐ClDA) and NaCl. After incubating the reaction mixtures for 4 h, the products (i.e., 5'‐FDA or SAM) were detected and quantified by HPLC. Under these assay conditions, we observed that each of the variants could catalyse the formation of 5'‐FDA from SAM and F–, although both the FlA‐K237E/N287P and FlA‐242_243InsLS mutants displayed considerably lower activity than wild‐type FlA (Fig. 2). In the former case (i.e., combined point mutations), the residual activity was almost negligible, whereas FlA‐242_243InsLS had a approximately 80% reduction in the catalytic output as compared to the wild‐type fluorinase. SAM formation from 5'‐FDA, on the other hand, was impaired for the FlA‐K237E, FlA‐K237E/N287P and FlA‐242_243InsLS variants. The FlA‐D241A mutant, in turn, retained the activity of the wild‐type enzyme in both reaction directions (Fig. 2), indicating a high degree of reversibility in its catalytic activity. Based on these results, the FlA‐K237E variant emerged as the most interesting one to explore trimer‐based fluorination. The next question was to analyse the actual oligomeric state of these fluorinase variants.

Table 1

Mutant FlA variants designed in this study.

FlA mutant	Design rationale
K237E	Disruption of the trimer:trimer hydrogen bond to E247
D241A	Disruption of the trimer:trimer hydrogen bond to Y286
K237E/D241A	Disruption of both hydrogen bonds above
242_243InsLS	Extending the trimer:trimer interface loop with residues from chlorinase
K237E/N287P	N287P disrupts backbone interaction with the other trimer; 287 is close to itself
K237E/I263P	I263P mimics chlorinase structure
K237E/L244T	L244T mimics chlorinase structure and disrupts hydrophobic packing; 244 is close to itself
K237E/I263P/L244T	Combination of all the rational mutations above

All mutant FlA were constructed by mutagenesis PCR using the flA gene from Streptomyces sp. strain MA37 as the template.

Fig. 2

Activity of rationally engineered FlA variants relative to the wild‐type (WT) FlA fluorinase. Each variant was tested both in the direction of 5'‐FDA (5'‐fluoro‐5'‐deoxyadenosine) formation (i.e., direct C–F bond formation) and in the reverse direction [i.e., S‐adenosyl‐l‐methionine (SAM) formation from 5'‐chloro‐5'‐deoxyadenosine]. Results represent average values of the relative enzymatic activity and the error bars indicate standard deviations from at least three independent biological replicates.

Biophysical exploration of the oligomerization state of FlA and its engineered variants

The oligomerization state of proteins is critical to their function, forming the basis for substrate recognition and underpinning the biophysical phenomena of allostery and cooperativity (Goodsell and Olson, 2000; Liu et al., 2020). Understanding the mechanism and pathways by which proteins assemble into higher order structures is critical both from a fundamental and applied point of view. Several biochemical and biophysical methods are available to examine the oligomerization behaviour of proteins. To study the oligomerization state of FlA and the set of mutants constructed herein, we resorted to both size‐exclusion chromatography (SEC) and small‐angle X‐ray scattering (SAXS) methodologies (Gell et al., 2012; Blanchet and Svergun, 2013).

The FlA variants were firstly subjected to SEC analysis (Fig. 3A). The purified FlA eluted as two distinct peaks of different intensity. We interpreted the appearance of the first and predominant peak to represent a hexameric population (within the 12–16 ml range), while the second peak – which has not been described previously – appears to be a trimeric subpopulation. In this case, approximately 75% of the purified protein sample was found to correspond to the hexameric population, as determined by the area under the curve (Fig. 3A). To substantiate this observation, we subjected the wild‐type FlA to SEC coupled to SAXS (Fig. 4A). Guinier analysis (Zheng and Best, 2018) of the scattering curve resulting from the major peak gave a radius of gyration (R g, in Å) and maximum distance (D max, in Å) fully compatible with a hexamer (Fig. 3B). A poor fit to the theoretical scattering curve in the mid‐s (scattering) range [evaluated as the squared Chi value (χ2) of the fitting curve, Fig. 4A] indicates some deviation in the overall shape of the wild‐type FlA enzyme with respect to the reported crystal structure (Dong et al., 2004). The minor peak from wild‐type FlA, however, did not contain enough material to obtain a meaningful scattering curve.

Fig. 3

A. Analysis of wild‐type (WT) FlA and rationally engineered mutants by means of size‐exclusion chromatography (SEC). Absorption values are indicated as mAU, arbitrary intensity units×10–3.

B. Experimentally determined radius of gyration (R g) and maximum size (D max) of the WT FlA and the FlA‐K237E variant deduced from small‐angle X‐ray scattering (SAXS) data (see Fig. 4). R g and D max values for a pure hexamer and trimer are likewise included in the table for comparison.

Fig. 4

Small‐angle X‐ray scattering (SAXS) profiles for (A) the wild‐type (WT) fluorinase and (B) the trimeric variant FlA‐K237E. The plots represent the logarithm of the scattering intensity (I, arbitrary units) as a function of momentum transfer (s, in Å–1), along with the best adjustment that fits to the experimental data according to the two possible oligomerization states. Squared Chi values (χ2) for these mathematical adjustments are included for both the hexamer and the trimer model.

In contrast with wild‐type FlA, all four engineered variants gave rise to a predominant peak in SEC experiments corresponding to the assumed trimeric population (located around 14 ml, Fig. 3A). We leveraged this feature to obtain a scattering curve from the trimer peak (Fig. 4B). Guinier analysis of scattering data from the major peak observed for the FlA‐K237E variant indeed fitted well with a proposed trimer (Fig. 3B). Again, a poor fit to the theoretical scattering curve in the mid‐s range indicated some shape deviation in solution from the reported crystal structure. In addition to the predominant trimeric population, all variants displayed low amounts of high molecular weight species or potential aggregates. Furthermore, preparations of FlA‐D241A had a significant level (approximately 33%) of hexamer, indicating an only partial disruption of this predominant oligomeric form – a feature that may correlate with the highly reversible activity profile observed for this variant (Fig. 2). Samples of the FlA‐K237E, FlA‐K237E/N287P and FlA‐242_243InsLS variants showed a single peak compatible with the trimeric form of the fluorinase. Prompted by these results, we selected the FlA‐K237E variant, which had both the highest activity in the forward reaction (Fig. 2) and showed a SEC‐SAXS profile compatible with a pure trimer (Figs 3 and 4), for further analysis of kinetic parameters.

Kinetic properties of the rationally engineered FlA mutants

As indicated above, the FlA‐K237E fluorinase variant was almost exclusively found as a trimer in solution (approximately 97%) yet it retained >80% of the wild‐type activity in assays carried out in the presence of SAM and F– (Fig. 2). A full kinetic characterization demonstrated that FlA‐K237E had a k cat of 0.20 ± 0.01 min–1 , very similar to the wild‐type enzyme (Table 2). By contrast, the Michaelis constant (K M) was higher (206.3 μM) than that of the wild‐type FlA from Streptomyces sp. strain MA37 (3.7 μM). This experimental determination of kinetic parameters resulted in a catalytic efficiency (k cat/K M ratio) for the FlA‐K237E fluorinase variant of 1 mM–1 min–1. Together with the unaffected k cat value and considering that the enzyme catalyses a single‐step reaction, the observed increase in the Michaelis constant [defined as K M = (k –1 + k cat)/k 1] could indicate increased dissociation of the substrate from the enzyme active site (represented by the k –1 component of K M). The location of the residue at position 237 in the trimer:trimer interface, distal to the active site (Fig. 1B), suggests a long‐range effect – plausibly resulting from increased flexibility of the trimeric form. Building on these experimental observations for the purified variant, we decided to explore if a trimer FlA mutant could be implemented for organofluorines biosynthesis as indicated in the next section.

Table 2

Kinetic constants of wild‐type (WT) FlA and the engineered FlA‐K237E variant.

Fluorinase variant	K M (μM)	k cat (min–1)	k cat/K M (mM–1 min–1)
FlA (WT)	3.7 ± 0.7	0.22 ± 0.01	58
FlA‐K237E	206.3 ± 34.1	0.20 ± 0.01	1

Kinetic constants were obtained from in vitro assays carried with purified enzymes. K M and k cat values are presented as averages ± standard deviation from at least triplicate determinations.

In vitro synthesis of fluorometabolites using a trimeric fluorinase

The successful engineering of a stable trimer (FlA‐K237E) variant with a similar k cat to the wild‐type prompted us to explore this fluorinase for fluorometabolite synthesis in a biotransformation setup. To this end, cell‐free extracts were prepared from cultures of E. coli BL21 transformed with plasmids encoding either FlA or FlA‐K237E, where the expression of the cognate flA genes is driven by the T7 RNA polymerase (Fig. 5A). The resulting cell‐free extracts were incubated in the presence of NaF and SAM for 20 h. After this incubation, the extracts were subjected to 19F‐NMR analysis and all fluorometabolites present in the samples were scanned. Figure 5B indicates that 5'‐FDA was the main fluorometabolite in cell‐free biotransformation experiments containing either FlA or FlA‐K237E. As expected, no fluorinated species were detected in control conditions (i.e., cells transformed with an empty vector). Interestingly, we observed secondary peaks in the 19F‐NMR spectra (with a weak intensity), which likely correspond to ribose‐based fluorometabolites (e.g., 5'‐fluoro‐5'‐deoxy‐d‐ribose 1‐phosphate). These fluorinated sugar derivatives could be potentially produced from 5'‐FDA by endogenous E. coli enzymes, yet the most abundant organofluorine in these reactions was 5'‐FDA. While further optimization would be needed to increase fluorometabolite titres using the trimeric FlA variant, these experiments indicate that the engineered FlA‐K237E variant can be used for biotechnological purposes towards organofluorine production.

Fig. 5

In vitro synthesis of fluorometabolites by wild‐type (WT) FlA and the rationally engineered FlA‐K237E variant.

A. Scheme of the plasmid constructs used for cell‐free biosynthesis of 5'‐fluoro‐5'‐deoxyadenosine (5'‐FDA), where the T7 RNA polymerase drives the expression of either flA (WT) or flA‐K237E, placed under control of the P T7 promoter as a chromosomally integrated module. Abbreviations are rep, replication module; and Km R, kanamycin‐resistance determinant.

B. High‐resolution 19F‐NMR profiles of the cell‐free extracts upon incubation with NaF and S‐adenosyl‐l‐methionine (SAM) as the substrate for FlA and FlA‐K237E. Empty indicates a control experiment where the E. coli cells used to prepare the cell‐free extract were transformed with the empty vector. Chemical shifts are expressed in parts per million, ppm.

Discussion

Nature has widely adopted halogenation for tuning the physicochemical properties of secondary metabolites in a wide variety of microbial species – mostly through the introduction of chlorine and bromine atoms (Fejzagic et al., 2019; Neugebauer et al., 2019; Adak and Moore, 2021). This occurrence is reflected in the rich and diverse enzymology for such biohalogenations that continues to be revealed and harnessed for biotechnology (Latham et al., 2018; Cros et al., 2022). However, this state of affairs hardly extends to fluorination. Biosynthesis of organofluorines remains a significant challenge to biochemistry due to the particular properties of the F– anion, which is a very poor nucleophile in water, and also because F has the highest oxidation potential of all of the halides (O'Hagan, 2008). These properties severely limit the potential for the creation of C–F bonds from F– by either nucleophilic or electrophilic processes in an aqueous environment, and only the nucleophilic fluorinase enzyme is known so far in Nature to execute this chemistry. Rationally manipulating this enzyme thus emerges as the way forward to enrich the biosynthetic landscape of organofluorine production.

Here, we demonstrated rational manipulation of the oligomerization status of FlA through selecting and implementing mutations at the dimer interface of the native dimer of trimers. We further established that SAM binding is compromised to some extent in the obligate trimer, indicative of a relationship between quaternary structure and substrate channelling (Sweetlove and Fernie, 2018), although variants such as FlA‐K237E still retained good capacity to synthesize fluorometabolites. From a broader perspective, engineering the oligomerization status of enzymes serves various purposes, both in fundamental and applied research (Liu et al., 2020). Forcing oligomerization is considered to be a viable strategy to improve enzyme function and stability under unfavourable environments (Yin et al., 2018; Jiang et al., 2019) – but this approach often complicates efforts in rationalizing functional enzyme characterization during fundamental research, which usually relies on assaying a mixture of oligomeric forms (Tong et al., 2005). A particularly relevant problem in biocatalysis using multimeric enzyme forms – in addition to the usually complex folding and assembling preparative processes involved – is the potential dissociation of enzyme subunits, which leads to deactivation and loss of activity, and even to contamination of the final product(s) of interest (Katchalski‐Katzir, 1993). In these cases, transforming homo‐oligomeric enzymes into functional individual monomers facilitates protein engineering efforts to match the demands of both functional characterization and industrial applications. By the same token, working with individual monomers enable studies about oligomeric structure‐activity relationships (Peverelli et al., 2016), and this general strategy has been applied in biomedicine for developing new drugs that target subunit interfaces of viral proteins (Sousa et al., 2011). Our present study not only provides evidence relevant for further functional studies of the fluorinase enzyme – an archetypal example of a homo‐oligomeric enzyme for which the hexameric form was supposed to be an essential feature needed for catalysis – but it also leads to practical applications in fluorine chemistry.

Trimeric fluorinase variants may be implemented towards fluorometabolite biosynthesis in synthetic enzyme cascades, which are increasingly being used for producing complex molecules through pathways that would be difficult to implement in an in vivo system (Zhou et al., 2021). Additionally, it seems plausible that loosening interactions between the enzyme and its substrate(s), as we have done here by oligomerization engineering, may expand the substrate range. This potential broadening in substrate specificity could enable fluorination of substrates beyond SAM, thereby providing a direct route towards the biosynthesis of fluorinated drug synthons that contain aromatic moieties attached to sugars (Inoue et al., 2020). Such building blocks are usually prepared by indirect fluorination techniques based on diazotization of anilines or exchange fluorination of activated haloaromatics and, more recently, by developing new fluorinating agents incorporated in somewhat complex synthetic methods (Cheng and Ritter, 2019). In addition, the emerging catalytic features of novel FlA variants can be also exploited in combination with robust microbial hosts (Volke et al., 2020a; Nikel et al., 2021) that support the synthesis of fluorometabolites, compounds typically known to be toxic for the cells (Bitzenhofer et al., 2021). Taken these results together, our present study highlights that the construction of bacterial cell factories for fluorochemicals production involving FlA will benefit from manipulating a variety of parameters – including the enzyme oligomerization status via direct, rational protein engineering – for optimal catalytic performance.

Experimental procedures

Plasmid design and construction and general molecular biology procedures

Mutations were designed based on the structure of fluorinase from Streptomyces sp. strain MA37 (Deng et al., 2014) and sequence alignments to other fluorinases and the SAM‐dependent SalL chlorinase (Eustáquio et al., 2008). Sequences were aligned and analysed using the Clustal Omega multiple sequence alignment tool (Larkin et al., 2007). The gene encoding FlA of Streptomyces sp. strain MA37 was obtained as a synthetic DNA fragment (Integrated DNA Technologies; Leuven, Belgium). The gene was cloned into a modified pET28(a)+ vector endowed with an N‐terminal His‐tag followed by a TEV (tobacco etch virus) cleavage site (GenScript; Piscataway, NJ, USA) with USER cloning. All primers were designed using the AMUSER tool (Genee et al., 2015). Phusion U DNA polymerase (Phusion U hot start master mix; ThermoFisher Scientific, Waltham, MA, USA) was used for PCR amplifications with the oligonucleotides listed in Table 3. PCR products were purified from a 1% (w/v) agarose gel and ligated with the USER enzyme to the receiving vectors using standard protocols (Sambrook and Russell, 2001; Ruiz et al., 2006; Volke et al., 2020b, 2021). Plasmids were transformed into chemically competent E. coli OneShot BL21(DE3) cells (ThermoFisher Scientific) for protein expression. Mutations were introduced in the flA sequence by PCR followed by USER cloning as described elsewhere (Volke et al., 2020c).

Table 3

Oligonucleotides used in this study.

Name	Oligonucleotide sequence (5'→3')	Direction
flA	AGGGCCAUGCCGCCAACGGCAGCCAG	Forward
flA	AACGGGCCUCCACGCGCACCTTCAGGCC	Reverse
pET28	AGGCCCGTUGATCGAGCACCACCACCACC	Forward
pET28	ATGGCCCUGAAAATAAAGATTCTCGCCGC	Reverse
K237E	ATTAGGGUGATGGTTCACGTAGTGGGCC	Forward
K237E	ATCTCCAGGUGTTTGCCCTGGCCGATGC	Reverse
K237E	ACCTGGAGAUCATCCTGGACGACGTGCT	Forward
K237E	ACCCTAAUCAAGTTTTTTGGGGTCGAGGT	Reverse
D241A	ATTAGGGUGATGGTTCACGTAGTGGGCC	Forward
D241A	AGCACGUCGGCCAGGATGATCTTCAGGTGTT	Reverse
D241A	ACGTGCUGCCGTTCGAAGCCCCGCT	Forward
D241A	ACCCTAAUCAAGTTTTTTGGGGTCGAGGTGCC	Reverse
D241A+K237E	ATTAGGGUGATGGTTCACGTAGTGGGCC	Forward
D241A+K237E	AGGATGAUCTCCAGGTGTTTGCCCTGGCC	Reverse
D241A+K237E	ATCATCCUGGCCGACGTGCTGCCGTTCGAA	Forward
D241A+K237E	ACCCTAAUCAAGTTTTTTGGGGTCGAGGTG	Reverse
InsLS	ATTAGGGUGATGGTTCACGTAGTGGG	Forward
InsLS	ACTGACAGGUCGTCCAGGATGATCTTCA	Reverse
InsLS	ACCTGTCAGUGCTGCCGTTCGAAGCCCC	Forward
InsLS	ACCCTAAUCAAGTTTTTTGGGGTCGAGGTGCC	Reverse
K237E+N287P	ACCTGGAGAUCATCCTGGACGACGTGCT	Forward
K237E+N287P	AGTGGGUACGGATAGGCCAGGCTGG	Reverse
K237E+N287P	ACCCACUGAAAGCCGGCCTGAAGGT	Forward
K237E+N287P	ACCCTAAUCAAGTTTTTTGGGGTCGAGGT	Reverse
K237E+N287P	ATTAGGGUGATGGTTCACGTAGTGGGCC	Forward
K237E+N287P	ATCTCCAGGUGTTTGCCCTGGCCGATGC	Reverse

Oligonucleotides are named according to the intended mutation(s) introduced in FlA. Sequences containing U residues were used for USER cloning.

Protein production in E. coli and protein purification protocols

A single transformant E. coli colony carrying the construct of interest was used to inoculate 5 ml of lysogeny broth (LB) medium supplemented with kanamycin (50 mg l‐1) and grown overnight at 37°C. Next, 400 ml of LB medium supplemented with kanamycin (50 mg l‐1) was inoculated with the overnight culture at a 1% (v/v) ratio and cultures were incubated at 37°C until the optical density measured at 600 nm (OD600) reached 0.5–0.8. At this point, protein production was induced by addition of 0.5 mM isopropyl‐β‐d‐1‐thiogalactopyranoside (IPTG; Sigma‐Aldrich Co., St. Louis, MO, USA). Cultures were then allowed to grow overnight at 18°C, after which the cells were harvested by centrifugation (4000 g, 15 min, 4°C). Cell pellets were stored at −20°C prior to protein extraction and purification.

Cells and proteins were kept on ice or in the cold room throughout the purification process. Bacterial pellets were dissolved in washing buffer (50 mM HEPES, pH = 7.4, 300 mM NaCl and 20 mM imidazole) supplemented with protein inhibitor cocktail tablets (cOmplete™, EDTA‐free protease inhibitor cocktail; Sigma‐Aldrich Co.). Cells were lysed by three passes through an Emulsiflex C5 high‐pressure homogenizer (Avestin Europe GmbH; Mannheim, Germany) and the lysate was cleared by centrifugation (12 000 g, 40 min, 4°C). The cleared lysate was incubated with a Ni‐NTA bead suspension (HisPur TM Ni‐NTA resin, ThermoFisher Scientific; 1 ml per construct) during 1 h with gentle shaking. Beads were washed twice with wash buffer and the bound protein eluted with an elution buffer containing 50 mM HEPES (pH = 7.4), 300 mM NaCl and 300 mM imidazole. TEV‐protease was added to the elution fractions (in a 1:20 fluorinase‐to‐TEV‐protease ratio) and the resulting fractions were dialyzed against 20 mM HEPES (pH = 7.8), 50 mM NaCl, 1 mM DTT and 0.5 mM EDTA overnight. Samples were incubated again with 0.7 ml Ni‐NTA beads (1 h) and the flow‐through was collected. Beads were washed with wash buffer as indicated above, and the washed protein fractions were combined with the corresponding flow‐through fractions. Protein purity was assessed using SDS‐PAGE (NuPAGE TM 4–12% Bis‐Tris protein gel; ThermoFisher Scientific). Finally, selected fractions were combined, and the buffer was exchanged to 20 mM HEPES (pH = 7.8), 150 mM NaCl and 3% (v/v) glycerol in centrifugal protein concentrators (10 000 Da cut‐off; Amicon Ultra; Sigma‐Aldrich Co.). Protein concentration was determined by assessing the absorbance at 280 nm in a NanoDropTM 2000 spectrophotometer (ThermoFisher Scientific). Protein fractions were aliquoted, flash‐frozen in liquid N2 and stored at –20°C until further analysis.

Samples for SAXS measurement were further purified by size‐exclusion chromatography using a SuperdexTM 200 increase 10/300 GL column (30 cm×10 mm, 8.6 μm particle size; Sigma‐Aldrich Co.) connected to an Äkta PURETM system (Cytiva; Global Life Sciences Solutions, Marlborough, MA, USA) and eluted with a buffer containing 20 mM HEPES (pH = 7.4), 150 mM NaCl and 3% (v/v) glycerol. Fractions containing pure enzyme were pooled, concentrated and stored at 4°C until further use.

Size‐exclusion chromatography

The oligomeric state of proteins was analysed using a SuperdexTM 200 increase 10/300 GL column (30 cm×10 mm, 8.6 μm particle size; Sigma‐Aldrich Co.) connected to an Äkta PURETM system. Proteins were diluted in the running buffer (50 mM HEPES, pH = 7.8 and 150 mM NaCl) and centrifuged (17 000 g, 10 min, 4°C) before applying the sample to the column. The particle size was estimated by comparing the elution volumes to a commercial molecular weight maker (Gel filtration standard; BioRad Laboratories Inc., Hercules, CA, USA).

In vitro enzyme activity assays

Enzyme activity assays were carried out with 5 μM purified FlA or mutants thereof, mixed with 300 μM SAM (New England Biolabs, Ipswich, MA, USA) [or, in some cases, 300 μM of 5'‐ClDA (Cayman Chemical Co., Ann Arbor, MI, USA)] and 1 mM l‐Met (Sigma‐Aldrich Co.) in a buffer containing 30 mM HEPES (pH = 7.8), 150 mM NaCl and 75 mM KF. Reactions were incubated at 37°C for 4 h, with 80‐μl samples taken at specific time points, incubated at 95°C for 5 min and centrifuged for 10 min at 17 000 g. The supernatants were analysed using an Eclipse Plus C18 column (100 × 4.6 mm, 3.5 μm particle size; Agilent Technologies, Santa Clara, CA, USA) connected to a HPLC system (DionexTM Ultimate 3000; ThermoFisher Scientific) operated at a flow rate of 1 ml min‐1. A gradient of solvent A [0.05% (v/v) acetic acid in water] and B (acetonitrile) was implemented as follows: 5‐12% (v/v) B in 1.5 min, 12% (v/v) B for 1 min, 12‐30% (v/v) B in 2 min and 30‐70% (v/v) B in 1.5 min. Enzyme kinetics were measured using 2 μM FlA or 1 μM mutants in a buffer containing 30 mM HEPES (pH = 7.8), 150 mM NaCl and 75 mM KF. The SAM concentration was varied between 0 and 800 μM, and samples were taken at specific time points. All samples were denatured by heating, and supernatants were analysed as explained above.

Data processing for (SEC)SAXS experiments

The ATSAS package was used for SAXS data processing and interpretation (Franke et al., 2017; Manalastas‐Cantos et al., 2021). All raw data were radially averaged and buffer‐subtracted automatically at the beamline. SEC‐SAXS data were processed with CHROMIXS (Panjkovich and Svergun, 2018); concentration series were merged in Primus (Konarev et al., 2003). All data were quality checked with the AutoRg (Petoukhov et al., 2007) and GNOM (Svergun et al., 1988) software packages. Finally, theoretical scattering curves were generated and fitted to experimental data with the CRYSOL tool (Franke et al., 2017).

Biosynthesis of fluorometabolites in bacterial cell‐free extracts

Cultures for production of wild‐type FlA and its FlA‐K237A derivative, as well as the corresponding cell‐free extracts for fluorometabolite synthesis, were prepared as described above with the following minor modifications. Bacterial cell pellets were dissolved in wash buffer (50 mM HEPES, pH = 7.4 and 300 mM NaCl) and lysed by three passes through an Emulsiflex C5 homogenizer (Avestin Europe GmbH). Cell lysates were incubated with 2.5 U ml‐1 of Pierce™ Universal Nuclease for cell lysis (ThermoFisher Scientific) for 30 min at room temperature under gentle shaking. The lysates were then centrifuged (10 000 g, 30 min, 4°C) and the corresponding supernatants were filtered (0.2 μm‐pore size); the recovered cleared lysates were immediately used for in vitro synthesis of 5'‐FDA. The in vitro, cell‐free production of fluorometabolites was carried out in 20 ml‐reactions containing 10 ml of cleared cell lysate, 5 mM NaF, 1 mM SAM and 50 mM potassium phosphate buffer (pH = 7.8) (Calero et al., 2020). Samples were statically incubated at 30°C for 20 h, and the reactions were stopped by heating at 95°C for 5 min, whereupon the resulting samples were stored at –20°C until further processing for fluorometabolite detection by NMR.

19F‐NMR analysis of fluorometabolites

Biological samples were lyophilized from a frozen solution in 50 mM potassium phosphate buffer using a Christ Alpha 1–2 LD Plus freeze dryer (Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany). To the freeze‐dried sample was added a MeOD/D2O solution (adjusted according to the sample size), and the resulting suspension was subjected to sonication to ensure dissolution of all fluorometabolites in the sample. After centrifugation (to remove any precipitate), 19F‐NMR experiments were recorded at 298 K on a Bruker AVANCE III HD instrument with either a SmartProbe BBFO+ (for proton coupled experiments) or a TCI CryoProbe (for proton decoupled experiments), using CFCl3 as an external reference (Calero et al., 2020; Wirth and Nikel, 2021).

Conflict of interest

None declared.

Author contributions

T.K., D.T., P.I.N., and D.H.W. were involved in conceptualization. T.K., P.T.L., P.C., and M.N.D. were involved in data curation. T.K., F.F., D.H.W., M.N.D., P.T.L., and D.O. were involved in formal analysis. T.K., D.W., P.I.N., and D.O. acquired funding. T.K., F.F., D.H.W., P.T.L., P.C., and M.N.D. were involved in investigation. T.K., D.H.W., and P.I.N. developed methodologies. T.K., D.H.W., D.T., and P.I.N. were involved in project administration. D.H.W., P.I.N., D.T., and D.O. carried out supervision. T.K. and D.H.W. were involved in writing original draft, and P.I.N. was involved in writing. All authors were involved in writing and final editing.