The introduction of fluorine into organic molecules can often improve their molecular stability and pharmacological properties ^{1, 2} . Organo-fluorines have been widely used in pharmaceuticals, diagnostics, agrochemicals, and materials ^{3, 4} , and it is estimated that 20–30% of commercial drugs, including many top-sellers, contain fluorine ⁵ . The increasing prevalence and success of organo-fluorines have instigated enormous efforts over the past decades in developing methodologies for efficiently introducing fluorine into organic molecules ^{5, 6} , however the unique properties of fluorine make these chemical incorporations challenging, usually needing harsh synthetic conditions and giving rise to moderate chemo- or stereo- selectivity ^{5, 6} .

In contrast to chemical synthesis, biocatalytic synthesis or biosynthesis provide better efficacy and selectivity under mild conditions, and can be easily scaled up for industrial production ⁷ . Although natural fluorination is rare, efforts in identifying new bio-fluorination machineries are continuing. Pioneered by O’Hagan and co-workers, the natural bio-fluorination pathway to produce fluoroacetate (FAc) and 4-fluorothreonine (4-FT) was biochemically characterized in the actinomycete Streptomyces cattleya ⁸ . Enzymatic fluorination in S. cattleya occurs via a fluorine-fixation step that is catalyzed by the unique fluorinase enzyme FlA through the nucleophilic attack of fluoride to SAM ^{9– 11} . The resulting fluoro-intermediate 5′-fluoro-5′-deoxyadenosine (5′-FDA) is further converted by five enzymes to give rise to the FAc and 4-FT ( Figure 1A and B) ⁸ . This powerful bio-fluorination machinery has been proved very successful in incorporating both ¹⁸F and ¹⁹F into nucleotides ^{12– 16} and in producing the anticancer drug candidate fluorosalinosporamide ¹⁷ . Very recently, its enormous potential for producing complex fluorinated molecules was demonstrated by Chang and co-workers ¹⁸ . Using an engaging biosynthetic strategy they were able to achieve site-specific incorporation of FAc into the polyketide backbone of 2-desmethyltriketide lactone. Merging this strategy with the FAC producing pathway, such as the fl pathway in vivo, could present opportunities to use living cells for the production of acetate-derived fluorinated natural products that include polyketides, fatty acids, terpenoids and steroids.

Synthetic biology approaches need pools of “genetic elements” to mix and match to build and optimize synthetic bio-pathways ¹⁹ . However, the only natural bio-fluorination tool available so far is the fl pathway. The limited natural bio-fluorination machinery forms the bottleneck for development of efficient synthetic bio-pathway to produce fluorinated natural products, thus an expansion of the genetic resources of biological fluorination is highly desired. Here, we report the identification of a latent fluorination pathway for FAc and 4-FT biosynthesis from the pathogen Nocardia brasiliensis ATCC 700358 and the biochemical characterization of a novel SAM-dependent fluorinase enzyme.

A putative gene ( nobA) encoding a protein with overall 79% identity to the fluorinase FlA was identified from the genome of N. brasiliensis ATCC 70035823 ( Figure 1A). By extending the searching region to the up-downstream of nobA, we further identified a gene cluster of 10 genes ( nobA-I, Figure 1A and Table 1). In this gene cluster, four catalytic genes ( nobA, B, D and M) encode the homologues of 5′-FDA synthase, 5′-FDA phosphorylase, 5′-FDRP isomerase and 4-FT transaldolase which are responsible for the conversion of fluoride to 5-FDRibulP and fluoroacetaldehyde to 4-FT in the fl pathway) ⁸ . Flanked those are genes for encoding auxiliary functions, including regulation ( nobC, F and G), transportation ( nobC, H) and SAM recycling ( nobE and I), among of which four ( nobF, G, I and H) are homologous to the counterparts in the fl biosynthetic pathway. The overall identity to the fl pathway genes suggested that the nobA cluster might be related to the 4-FT and FAc biosynthesis.

Besides the four catalytic genes described above, other two genes encoding an aldolase and a dehydrogenase to mediate the conversion of 5-FDRibulP to fluoroacetaldehyde and fluoroacetaldehyde to FAc are necessary. A homology search for the fluoroacetaldehyde dehydrogenase in the N. brasiliensis resulted in the identification of a gene with 78% identity ( nobJ, YP_006807765.1; Figure 1A and Table 1) from a remote site in the genome, suggesting that N. brasiliensis has the potential to produce FAc. Unlike the others, the gene encoding aldolases in the fl pathway is still elusive. Recently, four aldolase genes were identified from S. cattleya by genome sequencing and two of them were thought to encode the conversion of 5-FDRibulP to fluoroacetaldehyde ²⁴ . However, in the N. brasiliensis genome we could not find identical homologues by BLAST searching. Because this reaction can also be catalised by fuculose aldolase ⁸ enzyme, we searched for fuculose aldolase homologues, with no success. Instead, five putative aldolase genes ( YP_006810507.1, YP_006812798.1, YP_006805383.1, YP_006812725.1 and YP_006809408.1) were identified in N. brasiliensis based on the gene annotation in IMG database ( https://img.jgi.doe.gov/cgi-bin/w/main.cgi). We then compared the genetic sequences of the putative aldolases to their homologues in S. cattleya. Interestingly, two of these aldolase genes ( YP_004910624 and YP_004919742.1) which were previously not idenfied in S. cattleya shared sequence homology to the aldolase genes ( YP_006810507.1 and YP_006812798.1 respectively) of N. brasiliensis. Their homology (both have 62% identity) is in line with the average identity (58%) between the counterpart genes of nob and fl pathway, suggesting that these genes are probably involved in the FAc biosynthetic pathway.

Previous studies revealed the thioesterase FlK confers self-immunity to the FAc in the fl pathway ^{25, 26} . Inactivation of the flK gene does not affect cell growth in 2 M NaF ²⁷ , however can prevent production of fluorometabolites ²⁷ . Interestingly, we were not able to find the flK homologue in N. brasiliensis, although there are a few genes encoding thioesterases. To test the ability of N. brasiliensis to produce fluorometabolites, the bacteria were subjected to fermentation with a supplement of 2 mM NaF as a fluorine source ²⁸ . Despite the N. brasiliensis grows normally, no new fluorine signal in the culture except fluoride could be detected by F-NMR (data not shown). The lack of fluorometabolites might be caused by improper fermentation conditions or deficiency of the FlK homologue in the genome. In vivo gene manipulations including promoter activation and heterologous complementation of the flK in N. brasiliensis were also attempted, but failed due to the inaccessibility of genetic system.

Since the in vivo studies were inaccessible, we turned to in vitro characterization. The 5′-FDA synthase which catalyzes the unique fluorine-fixation reaction is one of the most interesting enzymes for the communities of enzymology and synthetic biology, thus we focused on the biochemical characterization of the NobA enzyme. The nobA gene was first codon optimized, cloned as an N-terminal 6×His tagged protein, and then expressed and purified from E. coli ( Figure 2A). In our experiment, the expression yield of NobA (7.9 mg/L) was lower than the yield of FlA (9.2 mg/L), and like FlA it natively occurs as a hexamer ( Figure 3) ¹⁰ . The fluorination activity of the recombinant protein was assessed in parallel with FlA which was utilized to produce the 5′-FDA as a control. After the reaction, the assay mixtures were quenched by 10% trichloroacetic acid (TCA) and subjected to HPLC analysis. A product strictly dependent on the addition of SAM, NaF and NobA was detected, and its identity was finally confirmed to be 5′-FDA by both HRMS and comparison to the product of FlA ( Figure 2B and C). These results, taken together, indicate that NobA, like FlA, can confer fluorine fixation, and that the nob pathway might be related to FAc and 4-FT biosynthesis.

With a new fluorinase in hand, we next compared the enzymatic properties of the protein. It has been previously reported that in the presence of L-amino acid oxidase (L-AAO), FlA can utilize Cl ^- but not Br ^- or I ^- to form the 5′-Chloro-5′-deoxyadenosine (5′-ClDA) ²² . Similarly, we observed that NobA was also able to perform this reaction with a similar reactivity ( Figure 4). The optimal pH of NobA was determined as 6.5, slightly lower than FlA at 7.0 ( Figure 5A). NobA shows no obvious effects on fluorination reactivity in the presence of metal ions, instead 1 mM Cu ²⁺ or Zn ²⁺ can severely inhibit its activity ( Figure 5B). The activity can be slightly increased by adding EDTA, suggesting the fluorination reaction is metal independent. The optimal temperature for fluorination reactivity is 37°C, however at this temperature SAM degradation is also expedited. In contrast, as 26°C can provide acceptable activity and slow down the degradation of SAM, this temperature was chosen throughout our kinetic studies. Although previous FlA kinetic studies have been reported ^{10, 29} , the data are inconsistent. Thereby, to precisely compare their reactivities, both NobA and FlA were kinetically measured in this study ( Figure 5C). The K _m values of NobA were 4153 μM for NaF and 416 μM for SAM, about twice the values of FlA (2167 μM for NaF and 210 μM for SAM). The turnover numbers ( k _cat ) of NobA were 0.073 min ^-1 for NaF and 0.139 min ^-1 for SAM, which were similar to the values of FlA (0.11 min ^-1 for NaF and 0.123 min ^-1 for SAM). The average k _cat / K _m of FlA (5.07×10 ^-5 μM ^-1 min ^-1 for NaF and 5.86×10 ^-4 μM ^-1 min ^-1) outnumbers 2.3 fold to the values of NobA (1.75×10 ^-5 μM ^-1 min ^-1 for NaF and 3.3×10 ^-4 μM ^-1 min ^-1 for SAM), and suggests that NobA is less efficient than FlA.

Similar to FlA, NobA retains the conservation of the unique 23-residue loop, which is related to site architecture and fluoride selectivity, and lacks the critical residues for SAM hydrolysis ( Figure S1) ^{30, 31} . This unique feature confirms that fluorinases form a distinct group of enzymes which differ from other SAM-binding members in the DUF62 enzyme family. The overall identity of the fluorinases is very high (79% identity), and only a region between the R192 to L202 (FlA residue numbering) is distinct ( Figure S2). Interestingly, in the crystal structure of FlA, this region constitutes a loop which links the N- and C-terminal domains to form an interface where the SAM and fluoride are bound ¹¹ . The variety of the loop region can probably influence the domain interaction and binding affinities of the substrates, and this might be one of the reasons for the retardation observed for NobA. Moreover, the residue S158 of FlA plays critical role in fluoride binding and desolvation ^{10, 32} . Interestingly, this residue also exists in the NobA ( Figure S2). To confirm the essentiality of S158 for NobA activity, this residue was further mutated into an alanine ( Figure 2A). The resulting recombinant protein NobA-S158A completely lost fluorination activity ( Figure 2B) and demonstrates it is essential for NobA activity. This observation also indicates that NobA shares an identical catalytic mechanism employed by NobA.

Actinomycetoma is a zoonotically chronic, granulomatous and subcutaneous tissue infection caused by actinomycetes that may often lead to amputation or death ³³ . N. brasiliensis is the major causative pathogen of the actinomycetoma infections in Mexico ^{23, 33} . Its pathogenic mechanism currently is unclear. FAc is a highly toxic poison (LD ₅₀ = 0.1 mg/kg mouse ≈ 1.2 μM) ³⁴ . In the human body it is first converted into the anabolite (-)-erythro-fluorocitrate (EFc), which exerts its toxicity by irreversibly inhibiting aconitase to cause cell death (IC ₅₀ = 0.01 μM) ^{35, 36} . Interestingly, our research revealed the N. brasiliensis has the genetic potential to produce FAc from fluoride. In human blood, the fluoride concentration is normally around 0.13 μM and can increase to 3.2 μM in some fluorine-polluted areas ³⁷ , which indicates N. brasiliensis can potentially produce FAc at levels that cause severe cytotoxicity and lead to development of the actinomycetoma.

In summary, we identified a latent pathway for FAc and 4-FT biosynthesis from the actinomycetoma pathogen N. brasiliensis ATCC 700358. Comparative analysis of the genomes of N. brasiliensis and S. cattleya revealed two common aldolase genes that were putatively involved in the FAc and 4-FT biosynthesis. By in vitro characterization, we confirmed a novel fluorinase, NobA, that can biosynthesize 5′-FDA from inorganic fluoride and SAM. This new fluorinase has similar substrate selectivity and characteristics to the homologue FlA, with a slightly less efficiency of reaction (2.3 fold). The lower reactivity might be attributed to the presence of a distinct loop region in the sequences of FlA and NobA, based on comparative sequence analysis. During the revision of this manuscript, O’Hagan and co-workers published the identification of the same fluorinase and two other fluorinases from S. sp MA37, Actinoplanes sp N902-109 ³⁸ . These fluorinases are highly conserved (76–79% identity to each other) but all have the distinct loop region which is identified by this study ( Figure S3). This feature indicates that fluorinases are a unique protein family whose evolution is also unusual, majorly concentrated at the interface of N- and C- domains. Moreover, by searching the homologue of the putative S. cattleya 5-FDRibulP aldolases ( YP_004910624 and YP_004919742.1, identified in this study) in the geome of A. sp N902-109 results identification of an aldolase (YP_007950157.1). It shares 65% identity to the aldolase from S. cattleya ( YP_004910624) and N. brasiliensis ( YP_006810507.1) indicating that this common homologue is likely responsible for the conversion of 5-FDRibulP. The discovery of new fluorinases and biosynthetic pathways increases the genetic resource of bio-fluorination and will benefit the future development of synthetic bio-pathways to produce fluorinated natural products.

The data referenced by this article are under copyright with the following copyright statement: Copyright: � 2014 Wang Y et al.

Data associated with the article are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Data sets showing the fluorinase catalysed conversions of fluoride and chloride to 5′-FDA and 5′-ClDA are publicly available in ZENODO.

ZENODO: Conversion of fluoride and chloride catalysed by SAM-dependent fluorinase in Nocardia brasiliensis, doi: 10.5281/zenodo.8339 ³⁸ .