Cofactor Trapping, a New Method To Produce Flavin Mononucleotide

Abstract

We have purified flavin mononucleotide (FMN) from a flavoprotein-overexpressing Escherichia coli strain by cofactor trapping. This approach uses an overexpressed flavoprotein to trap FMN, which is thus removed from the cascade regulating FMN production in E. coli. This, in turn, allows the isolation of highly pure FMN.

Flavins (Fig. 1) are ubiquitous in nature and take part in many biochemical reactions mainly in the form of the coenzymes flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD), which function as redox cofactors in different flavoproteins. Riboflavin (RF) is the precursor of FMN and FAD and part of the vitamin B complex.

FIG. 1.

Chemical structures of different naturally occurring flavins.

RF has traditionally been synthesized chemically to be used for food and feed fortification. More recently, biotechnological processes employing various bacteria, yeasts, and fungi have become commercially competitive and are increasingly replacing chemical synthesis (18). In contrast, the phosphorylated flavin derivates FMN and FAD are still mostly produced by chemical synthesis, which is performed via phosphorylation of RF using phosphoroxychloride (POCl₃) in γ-butyrolactone as the solvent. Tedious and expensive downstream processing is needed to separate the respective phosphorylated flavins from unreacted RF and reaction side products (2, 6). Flavins are important as coenzymes for many biological reactions in vivo and are thus needed for biochemical reactions as well as for medical applications (4, 16). Hence, it is of great interest to develop a biological system suitable for overproducing and purifying phosphorylated flavins (e.g., FMN and FAD) directly from a microbial host. Thus, metabolic engineering has been used in recent years to obtain FMN-overproducing strains of Escherichia, Enterobacter, or Bacillus spp. (3, 9, 14). Usually, those strains overexpress a bifunctional flavokinase/FMN adenylyltransferase and require the addition of large amounts of expensive ATP and RF to the culture supernatants (15) to overcome RF availability limitations due to the repression of RF biosynthesis by FMN. Presently, riboflavin biosynthesis is best studied in Bacillus subtilis, where FMN represses biosynthesis (11, 14) via a feedback mechanism involving an FMN-binding riboswitch, or so-called RFN element, found in the 5′ untranslated region (UTR) of the RF biosynthesis operon (17, 21). The RFN element either initiates transcriptional termination or sequesters an adjacent ribosome-binding site in the presence of FMN (21). Comparative genome analyses have identified RFN elements in the 5′ UTR of several RF biosynthesis genes in different proteobacteria, including, among others, Escherichia coli (20); however, direct experimental evidence is lacking. Hence, metabolic engineering of bacterial RF overproducers such as B. subtilis with the aim to selectively upregulate FMN production without influencing RF biosynthesis appears to be difficult.

Recently, also the use of engineered flavogenic yeast strains has been reported for the overproduction of FMN. An RF-overproducing Candida famata strain with multicopy chromosomal insertions of the FMN1 gene encoding the RF kinase of Debaryomyces hansenii produced 200 to 250 mg/liter FMN in culture supernatants under optimized high-density cultivation conditions (22). However, FMN production by the engineered strains was accompanied by accumulation of large amounts of RF in the medium (22), thus requiring further tedious downstream processing procedures (2, 6, 7). Apparently, an easy and commercially feasible method for the biological production of highly pure FMN is currently not available.

We have developed a novel biological production system yielding highly pure FMN by using cofactor trapping. This strategy does not rely on metabolically or otherwise engineered FMN overproducers but instead uses a standard Escherichia coli strain expressing an FMN-binding flavoprotein which traps most of the cellular FMN within the protein, thereby removing FMN from the regulatory cascade regulating biosynthesis. In biocatalytic production processes, in situ product removal (ISPR) has long since been employed to remove inhibitory or toxic products (5, 13). Such in vitro strategies generally rely on process engineering techniques using physical adsorption, liquid-liquid extraction, or distillation (13) to remove the product from the bioreactor during biotransformation. In contrast, the cofactor-trapping method represents an in vivo approach where the product FMN is sequestered and stored within the cell during fermentation.

To obtain the final product FMN, we have overexpressed an engineered derivative of a naturally occurring flavoprotein, namely, the Pseudomonas putida blue-light photoreceptor protein PpSB2-LOV (8, 10), which belongs to the light, oxygen, voltage (LOV) family (1). In LOV domains, in the dark, FMN is noncovalently bound within the protein. An intricate network of conserved amino acids surrounding the flavin isoalloxazine ring, the ribityl chain, and the FMN phosphate moiety stabilizes the binding of the flavin chromophore within the protein (12). Wild-type PpSB2-LOV accepts both FMN and RF as chromophores at a ratio of 70%:30% (8). We had previously engineered this protein to exclusively bind FMN in stoichiometric amounts (8) and therefore assumed that overproduction of this flavoprotein in E. coli should result in the trapping of all intracellular available FMN molecules not otherwise bound. As a consequence, FMN molecules would be removed from the regulatory cascade that represses RF and hence FMN overproduction. After protein overexpression, the bound FMN can be released from the flavoprotein and purified from crude cell extracts.

The method comprises protein overexpression using an autoinduction system described previously (8, 19). In brief, the medium contained 12 g/liter casein hydrolysate, 24 g/liter yeast extract, and 5 g/liter glycerol, in 100 mM potassium phosphate buffer, pH 7.0. The medium was supplemented for induction with 0.5 g/liter glucose and 2 g/liter lactose, respectively. Cells were grown for 16 h at 37°C in a 10-liter batch fermentor under constant stirring at 500 rpm. Subsequently, cells were harvested and 6.5 g (wet weight) lysed by passing the cells four times through a chilled 40K French pressure cell (Thermo Scientific, Waltham, MA) at a constant pressure of 5.5 × 10⁷ Pa. During cell lysis, DNase I (Roche, Basel, Switzerland) was added. Cell debris was removed by centrifugation at 34,000 × g for 30 min. The flavoprotein-containing crude lysate was concentrated to about 1 ml using a Vivacell concentrator unit (molecular mass cutoff, 10 kDa). In order to remove small molecules from the lysate prior to protein denaturation, the concentrator unit was refilled once with 250 ml of Millipore H₂O and concentrated again to 1 ml. Subsequently, all cellular proteins, including the overexpressed flavoprotein, were denatured using cold perchloric acid, which results in the release of FMN from the overexpressed flavoprotein. Acid treatment was repeated until visible protein precipitation ceased to occur (usually 6 times). Precipitated protein and cell debris were removed by centrifugation, and the obtained extract was neutralized by adding 4 M potassium hydroxide. Subsequently, FMN was purified by ion-exchange chromatography (DEAE SepharoseFF; GE Healthcare, Buckinghamshire, United Kingdom) using for elution a stepwise increased concentration of HCl (10 mM, 15 mM, and 30 mM; 2 column volumes each). The eluted samples were lyophilized and resuspended in water prior to analysis for purity and yield. Electrospray ionization mass spectrometry (ESI-MS) and ¹H nuclear magnetic resonance (NMR) spectroscopy were used to assess the chemical identity and structural integrity of the samples.

Expression of the mutated LOV protein in E. coli BL21(DE3) from the strong promoter P_T7 results in a yield exceeding 150 mg per liter of culture, corresponding to a wet weight of cells of about 6.5 g (data not shown). Notably, this yield was achieved without any optimization of protein expression or cultivation conditions. A protein yield of 150 mg/liter and an optimal protein-to-FMN ratio of 1:1 would result in an estimated theoretical yield of 4.5 mg FMN per liter of culture, assuming the binding of one FMN molecule per subunit of the dimeric LOV protein. The above-described method yielded a yellow, highly fluorescent flavin sample for which thin-layer chromatographic analyses indicated FMN as the main flavin species (data not shown).

The quantity and quality of the purified FMN preparation were estimated employing UV/Vis spectrophotometry (Fig. 2) and high-performance liquid chromatography (HPLC) (Fig. 3) using a standard curve of known amounts of FMN as described previously (8). Repeatedly, a yield between 0.2 and 0.5 mg FMN per 6.5 g of E. coli cells, corresponding to 5 to 10% of the theoretical value, could be obtained. Hereby, the yield is crucially influenced by the quality of the flavoprotein-expressing cells. In all cases, freshly transformed cells used for biomass production resulted in a better FMN yield than did (long-term) stored cell pellets. Moreover, HPLC analysis demonstrated a purity of more than 95% for the preparation with no other flavin species present. Peak identity was verified on the basis of typical flavin spectra with maxima at 272 nm, 390 nm, and 450 nm.

FIG. 2.

UV/Vis spectra of FMN obtained by cofactor trapping from E. coli overexpression cultures (A) and commercially available FMN (Sigma-Aldrich, St. Louis, MO) of the highest purity available (B). Equal amounts of FMN were dissolved in fast protein liquid chromatography elution buffer (30 mM HCl). AU, arbitrary units.

FIG. 3.

(A) HPLC chromatogram of FMN obtained by cofactor trapping from E. coli overexpression cultures. The only flavin species present in the preparation obtained from cofactor trapping elutes with a peak retention time identical to that of authentic FMN. Apart from minor traces of RF (>1%), no other flavin species could be detected. (B) HPLC chromatogram of commercially available FAD, FMN, and RF (Sigma-Aldrich, St. Louis, MO; highest purity available). AU, arbitrary units.

The structural integrity and chemical identity of the obtained FMN preparation were further analyzed by MS analyses and by ¹H NMR. Both methods independently confirmed the identity of the preparation as FMN (data not shown).

Our results clearly demonstrate that cofactor trapping represents an easy and inexpensive method for the biological production of highly pure FMN, a phosphorylated metabolite which is currently commercially available only by chemical synthesis involving tedious downstream processing steps. This method does not rely on engineered flavogenic yeasts or other flavin-overproducing microorganisms but instead uses a well-established E. coli overexpression system. The obtained yields of about 0.2 to 0.5 mg FMN are presently low. Theoretically possible reasons might include (i) the presence of excess apoprotein (i.e., the flavoprotein without the flavin chromophore) which may compromise the overall yield or (ii) significant losses during purification. Thus, an optimization of expression conditions (medium composition, growth temperature, and induction method) may be required to achieve a more favorable protein-to-chromophore ratio without compromising protein expression levels and thus could help to increase the overall FMN yield. However, when a small amount of the flavoprotein was purified from the fermentor-grown cells, apoprotein content was estimated to be around 10 to 20%. Thus, apoprotein content compromises the overall FMN yield to some extent but cannot account solely for the low yield. Thus, most importantly, optimization of the purification procedure should result in minimization of FMN loss during downstream processing.

Despite the current limitations, we believe that the general concept of cofactor trapping that we present here exemplified by FMN, a biologically difficult-to-produce metabolite, could also be employed for the biological production of a wide variety of different metabolites. This method may prove especially useful in situations where product feedback inhibition limits (i.e., represses) the production efficiency of the entire biosynthetic pathway. To this end, respective binding proteins can be identified or existing binding proteins can be optimized by directed evolution or rational protein engineering to create a toolbox of biological traps useful for metabolite production and purification.