INTRODUCTION
The processing of fish for human consumption gives rise to 50 to 70% by-products, such as heads, frames, viscera, blood, and trimmings (
1), leading to an estimated 1 million tons of by-products annually in Norway alone (
2). In recent decades, there has been great progress in the development of enzyme-based processes by which fish by-products are converted into marine ingredients, such as peptides and oils. Today, these ingredients are mainly used to manufacture animal and pet feed, but since the by-products contain high-quality proteins, they also have a great potential for human consumption. Depending on processing conditions, however, enzymatically derived fish protein hydrolysates may suffer from a distinct malodor (
3), which is described as an unpleasant smell associated with rotting fish. This currently limits their potential in the human consumption market.
Trimethylamine (TMA) is a major contributor of malodor from fish and is recognizable as a pungent fish odor (
4). Even very low TMA levels, down to 0.00021 ppm, have been reported as recognizable by humans (
5). In fish, TMA accumulates postmortem as a result of bacterial conversion of the oxygenated odorless precursor, trimethylamine
N-oxide (TMAO) (
6). TMAO acts as an osmolyte that stabilizes proteins during environmental stress, such as osmotic pressure, and it is therefore believed to be particularly abundant in species located in the depths of the ocean (
7). As TMA is common in marine environments, a possible enzymatic solution to this malodor challenge may be found in marine bacteria that can utilize TMA as their sole carbon and/or nitrogen source, where the first step is to convert TMA to odorless TMAO (
8).
The oxidation of TMA to TMAO is catalyzed by monooxygenases, which, as reflected by their name, are enzymes that catalyze the insertion of a single oxygen atom into their substrates. Several classes of monooxygenases have been described (
9). Class B flavin-dependent monooxygenases include class I Baeyer Villiger monooxygenases (BVMOs),
N-hydroxylating monooxygenases (NHMOs), YUCCAs, and flavin-containing monooxygenases (FMOs), and together they are involved in several key biological processes (
10,
11). Structurally, class B flavin-dependent monooxygenases contain two Rossmann fold domains that harbor dinucleotide binding motifs, namely, for their tightly bound flavin adenine dinucleotide (FAD) and NADP (NADPH) cofactors, on which the enzymes depend for their catalytic action (
10). The TMA-oxidizing monooxygenases are classified as FMOs, and both mammalian and bacterial FMOs appear to utilize a “cocked-gun” mechanism (
12,
13), where FAD is reduced by NADPH in the absence of substrate, generating the reactive intermediate C4a-hydroperoxy-FAD and NADP
+. The C4a-hydroperoxy-FAD intermediate is stabilized and shielded by NADP
+ (
13,
14) and readily inserts one oxygen atom into the substrate upon entering the active site. Whether the NADP
+ remains bound to the active site during catalysis has been debated (
13–15).
FMO enzymes that convert TMA to TMAO have been named trimethylamine monooxygenases (Tmms) (
16); they are promiscuous enzymes, found in all kingdoms of life (
12,
16–18), that can oxidize a range of additional nonpolar substrates, such as indole and methimazole (
13,
16). The most studied Tmm is the human hepatic enzyme FMO3, which is responsible for TMA-to-TMAO conversion in the liver (
19). Malfunction of FMO3 is linked to the metabolic disorder trimethylaminuria, or fish odor syndrome. Patients suffering from this disease are unable to rid their body of TMA, leading to a strong bodily odor resembling that of spoiled fish (
20).
Tmms have been found to be abundant in marine bacterial metagenomes, suggesting that they play a crucial role in the carbon and nitrogen cycling in the ocean (
16). The first bacterial Tmm, mFMO, was discovered in
Methylophaga aminisulfidivorans strain SK1, which is marine and belongs to the
Gammaproteobacteria (
13,
14,
18). Other marine bacterial Tmms shown to be active have been found in the
Alphaproteobacteria and include RnTmm from
Roseovarius nubinhibens strain ISM (
15) and Tmms from
Methylocella silvestris strain BL2,
Roseovarius sp. strain 217,
Ruegeria pomeroyi strain DSS-3,
Pelagibacter ubique strain HTCC1002, and
Pelagibacter ubique strain HTCC721 (
16). Tmms have also been described in other habitats, including NiFMO from
Nitrincola lacisaponensis, which was isolated from an alkaline saline lake (
21), and cFMO from
Corynebacterium glutamicum, isolated from soil (
22). The fact that
C. glutamicum belongs to the
Actinobacteria suggests that there may be Tmms of potential industrial interest outside the
Proteobacteria.
The aim of this study was to identify bacterial Tmms that efficiently convert TMA to TMAO when expressed recombinantly in Escherichia coli and to test their ability to perform this catalytic reaction in a TMA-rich salmon protein hydrolysate. To this end, a sequence similarity network (SSN) analysis was employed to select 45 Tmm candidates from six taxonomic groups. Successfully expressed and soluble candidates were purified and assessed for their ability to convert TMA to TMAO by monitoring both NADPH cofactor consumption and TMAO product formation. As a first step toward industrial application, the three best-performing Tmms were characterized biochemically, by studying their temperature and pH profiles, as well as by assessing their structural and functional stability. Finally, as a proof of concept, we demonstrate that our top three Tmms were all able to convert TMA to TMAO in a salmon protein hydrolysate.
DISCUSSION
In the present study, we aimed to identify Tmm enzymes capable of converting TMA to TMAO with high specific activity. Tmms may prove to be useful biocatalysts for transforming salmon protein hydrolysates into more desirable ingredients for human consumption, by oxidizing the malodorous TMA to the odor-neutral TMAO. The TMA levels in raw fish filets from Atlantic salmon stored at 4°C for zero to 2 days are reported to contain approximately 0.002 to 0.01 mg/g tissue (wet weight), which increases to 0.19 to 0.33 mg/g tissue after a week of storage at the same temperature (
31,
32). TMA levels vary in the biomass depending on fish tissue, species, temperature, and processing (
4,
31,
32). The protein hydrolysate made from Atlantic salmon applied in the present study had TMA levels of 1.6 mg/g hydrolysate (about 64.6% dry weight) (
Fig. 5A), more than 160-fold the amount found in short-term cold-stored fish filet. These data suggest that all or a major fraction of the water-soluble TMA is retained in the soluble fraction during enzymatic hydrolysis and gives rise to increased TMA levels in the protein hydrolysates, which further emphasizes the challenge of achieving a desirable organoleptic quality. Accepted threshold levels of TMA in Atlantic salmon are reported to be 0.295 mg/g tissue (wet weight) (
31), highlighting the need for novel strategies targeting this particular volatile odor molecule.
Previous attempts to solve problems with undesired odors of marine hydrolysates include odor masking. Salt-water clam hydrolysates have been treated with tea polyphenol (
33), which masked the odor by affecting TMA, and Atlantic salmon hydrolysates have been treated by adding sugars that proved effective, despite not affecting TMA (
34). The sugar-treated salmon hydrolysates, however, had a distinct grilled odor, which may not be desirable. In contrast to odor masking, we have shown that the Tmms, such as T10, T37, and T51, can be used to remove TMA from such hydrolysates (
Fig. 5A), possibly avoiding unwanted side effects.
In the proof-of-concept experiment performed on a salmon hydrolysate, enzymatic activity against TMA was only observed after the cofactor NADPH was added to the hydrolysates. Other TMA-rich fish products may have sufficient intrinsic levels of NADPH to drive Tmm enzymatic activity, but where that is not the case, cofactor dependency must be addressed. NADPH is an expensive chemical, and hence, adding NADPH to an industrial process is highly unlikely to be cost effective. One solution is to engineer the enzymes to accept alternative cofactors (
35), which may be present in the TMA-rich biomaterial of interest or are less expensive to add. Another solution is to regenerate the cofactor (
36,
37). Flavin binding monooxygenases, such as class I type I BVMOs, have been fused to a glucose dehydrogenase to obtain a continuous oxidation reaction driven by the dehydrogenase-catalyzed regeneration of NADPH from NADP
+ (
38).
The three Tmms that were subjected to biochemical characterization have different temperature and pH profiles and can thus serve as unique starting points for enzyme engineering to tailor the enzymes for industrial applications. Among our panel of enzymes, the mFMO (T10) from
M. aminisulfidivorans may be the best candidate for the process used to generate the salmon hydrolysate studied here, which operates at 40 to 60°C and pH 6.0 to 7.0 (
3,
29). Still, engineering is likely to be required to increase the enzyme stability and temperature optimum. In contrast, for processes that operate at lower temperatures, T37 from
N. alba and T51 from
M. terreus are likely to be more suited and better starting points for engineering, if required. Finally, the Tmms identified here demonstrate that there are Tmms in several taxonomic groups, including
Actinobacteria,
Bacteroidetes,
Gammaproteobacteria, and
Alphaproteobacteria, that could be further explored for their industrial applicability.
MATERIALS AND METHODS
SSN construction.
A sequence similarity network (SSN) was created for the InterPro flavin monooxygenase-like (FMO-like) family (InterPro accession number IPR020946) using the Enzyme Function Initiative Enzyme Similarity Tool (EST-EFI) (
23). The SSN was constructed using UniRef90 and default settings, and the SSN was finalized using an alignment score threshold of 35 and only including sequences with a length of 100 to 800 amino acids. Due to the large size of the network, a representative network, where each node is a collection of proteins with 40% or higher identity, was downloaded from the server and processed further using Cytoscape 3.8.0 (
39). Cytoscape was used to remove nodes containing only nonbacterial sequences and edges with alignment scores lower than 50 and to visualize the final SSN using the yFiles organic layout.
MSA and phylogenetic tree construction.
A multiple-sequence alignment (MSA) of the Tmm candidate sequences was made using Clustal X (
40) with default parameters. The alignment was visualized with JalView 2.11.1.0 (
41), and sequence logos were made using WebLogo (
42). The MSA was used to construct a phylogenetic tree with MrBayes version 3.2.6 (Ronquist et al., 2012), using the following settings: the priors for the amino acid model were set to mixed, and the number of generations used was 200,000. The resulting phylogenetic tree was visualized with FigTree version 1.1.3, and the sequences from the SSN C5 cluster were used as an outgroup to root the tree.
Molecular cloning of candidate enzymes.
All candidate enzymes were ordered as DNA from Twist Bioscience (San Francisco, CA, USA) and subcloned and sequenced as previously described (
43,
44). Briefly, codon-optimized genes flanked by SapI sites were subcloned by fragment exchange (FX) cloning into the pBXC3H vector containing a C-terminal 6×His tag. The ligation reaction product was transformed into
E. coli MC1061 cells, and clones were selected on LB agar (1% [wt/vol] tryptone, 0.5% [wt/vol] yeast extract, 1% [wt/vol] NaCl, 1.5% [wt/vol] agar-agar) supplemented with kanamycin (50 μg/ml; Sigma-Aldrich). Plasmids were isolated using the NucleoSpin plasmid purification kit (Macherey-Nagel), and Sanger sequencing was used to confirm correct cloning.
Expression and solubility screening.
All enzyme constructs were expressed in E. coli MC1061 cells in 24-well deep-well plates in LB medium (Sigma-Aldrich) supplemented with 0.1% (wt/vol) l-tryptophan (Sigma-Aldrich) and 100 μg/ml ampicillin at both 20°C and 30°C for 16 h after induction with l-arabinose (Sigma-Aldrich) to a final concentration of 1% (wt/vol). Proteins were purified using Ni-NTA spin columns (Qiagen), following the Qiagen protocol. Cells were collected by spinning at 4,000 × g for 30 min at 4°C and lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 1 mM phenylmethanesulfonyl fluoride [Sigma-Aldrich], 0.1% n-dodecyl β-d-maltoside [Sigma-Aldrich], 5 μg/ml DNase [Sigma-Aldrich], 0.2 mg/ml lysozyme [Sigma-Aldrich]). Columns were equilibrated with equilibration buffer (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 10 mM imidazole) before incubation with cell lysate and washed with wash buffer (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 30 mM imidazole). Proteins were eluted with elution buffer (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 500 mM imidazole). Total protein fractions of cell lysate, soluble fractions, and eluted proteins were analyzed by SDS-PAGE (Bio-Rad) (Fig. S1 in the supplemental material).
Medium-scale protein expression and purification.
Soluble candidate enzymes were expressed in E. coli MC1061 cells in 200 ml LB medium supplemented with 0.1% (wt/vol) l-tryptophan at 20°C or 30°C for 16 h after induction with l-arabinose to a final concentration of 1% (wt/vol). All subsequent steps were carried out at 4°C. Cells were collected by centrifugation at 19,000 rpm for 20 min. The cell pellets were resuspended in lysis buffer and subjected to three freeze-thaw cycles, including 10 min on dry ice and ethyl alcohol (EtOH), followed by 15 min in a room temperature water bath. Following freeze-thawing, the cell lysate was sonicated briefly 3 times for 10 s at 60% amplitude to remove DNA. The cell lysate was finally spun at 17,000 rpm for 30 min, and the cleared lysate was passed over preequilibrated prepacked Ni-NTA columns (Thermo Fisher). The resin was washed with 10 column volumes of wash buffer before elution by flowing elution buffer over the Ni-NTA columns. The eluted protein was then subjected to buffer exchange using equilibrated (50 mM Tris-HCl, pH 7.5, 100 mM NaCl) PD10 columns (GE Healthcare). Protein concentrations were measured using the Bradford method. Enzyme purity was checked by SDS-PAGE (Fig. S1). The eluted proteins were stored with 10% glycerol at −20°C until further use.
NADPH assay.
Activity assessments were conducted by detecting the conversion of NADPH to NAD
+ at 340 nm in 96-well plates. All assays were carried out in reaction buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl) supplemented with 0.25 to 0.5 mM NADPH (Sigma-Aldrich) and 0.01 to 0.02 mg/ml enzyme. Reactions were initiated by the addition of 1 mM TMA (Sigma-Aldrich). We assumed the concentrations of TMA and NADPH to be saturating (>10×
Km) for all enzymes to be tested based on previous publications on T10 (
Km, TMA = 7.3 μM and
Km, NADPH = 13.0 μM) (
13) and T72 (
Km, TMA = 45.6 μM and
Km, NADPH = 8.2 μM) (
21). The decrease in absorbance at 340 nm was measured continuously over 30 min by UV-visible (UV-Vis) spectrophotometry in a Synergy HT multi-mode microplate reader. The initial rates were determined from linear fits of the absorbance versus time and corrected by the results for the blank. All assays (performed in triplicates) were performed at 21°C unless stated otherwise. In all cases, 1 unit (U) of enzyme activity was defined as the amount of enzyme required to transform 1 μmol substrate in 1 min under the assay conditions using the reported extinction coefficient (ε
340 = 6.22 mM
−1 cm
−1 for NADPH).
pH optimum and stability.
pH optimum and stability were measured by the NADPH assays as described above, using Britton-Robinson buffer (pH 5.0 to 9.0) (
45) instead of reaction buffer. For stability assays, enzymes were diluted to 0.5 mg/ml in the various buffers and incubated at 20°C for 2 h before measuring residual enzymatic activity as described above. Relative pH stability was plotted (GraphPad Prism 8) by setting the optimum of each enzyme as 100%.
Temperature optimum.
The temperature optimum was measured using indirect enzymatic assays as described above but using 2-ml cuvettes and an Agilent 8453 UV-Vis spectrophotometer connected to a circulating water bath. Specific enzyme activities were determined at temperatures from 20°C to 60°C in intervals of 5°C. Reaction mixtures contained temperature-equilibrated reaction buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl), 0.25 mM NADPH, and 1 mM TMA. Reactions were started by adding 0.01 to 0.02 mg/ml purified enzyme and continuously monitored at 340 nm over 5 min.
Temperature stability.
Freshly purified enzymes were diluted to 1.0 mg/ml in reaction buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl) and the initial enzyme activity was measured as described above. Enzymes were next incubated for 1 h at temperatures from 30°C to 54°C using a PCR thermocycler machine (Bio-Rad). After incubation, samples were cooled for 10 min at 4°C, followed by 10 min at room temperature before spinning for 2 min using a small tabletop centrifuge. Residual enzyme activity was measured as described above and was plotted by setting the initial activity as 100%. The temperature where half of the activity was lost was found by 4-parameter logistic regression using GraphPad Prism8.
CD spectroscopy.
Melting temperatures were recorded for enzymes diluted to 0.6 mg/ml. Circular dichroism (CD) spectra were acquired between 190 and 270 nm with a Jasco J-720 spectropolarimeter equipped with a Peltier temperature controller, employing a 0.1-mm cell. Spectra were analyzed, and denaturation temperatures (Td) were determined at 220 nm, between 10 and 95°C, at a rate of 30°C per h in reaction buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl). Td (and standard deviation of the linear fit) was calculated by fitting the ellipticity (millidegrees [mdeg]) at 220 nm at each of the different temperatures using 4-parameter logistic regression (n = 3 technical replicates).
Capillary electrophoresis electrospray ionization time-of-flight mass spectroscopy (CE-ESI-TOF-MS) sample preparation.
Enzymatic reactions were set up in duplicates in 50-kDa filters in volumes of 0.3 ml containing assay buffer, 0.25 mM NADPH, and 5 to 10 ng/ml enzyme at 30°C and 500 rpm. Reactions were initiated by the addition of 1.0 mM TMA. The filters were spun at 14,000 × g for 2 min at room temperature to stop the reaction. The flowthrough was collected, diluted 1:1 with a mixture of H2O-acetonitrile (95:5), 0.2 M formic acid (CH2O2), and 0.4 mM methionine sulfone (MeSO4), and vortexed for 5 min. Samples were stored at −80°C until analysis. To quantify the TMA and TMAO, two series of calibration samples were prepared by spiking different quantities of TMA (concentration range in the final sample, 10, 25, 50, and 100 mg/liter) and TMAO (0, 5, 10, 25, and 50 mg/liter).
Salmon protein hydrolysate (64.4% dry weight; provided by Biomega Group) was produced from fresh salmon by-products processed with proteases. The hydrolysate was diluted 1:5 (wt/vol) in ultrapure water due to its viscous nature in order to be able to analyze the solution using CE-ESI-TOF-MS. Diluted hydrolysate was sonicated in a water bath at 50 Hz for 5 min and vortexed for 5 min before centrifugation at 16,000 × g for 10 min and collection of the supernatant (SPN). The pH of the SPN was 6.10. Samples were prepared with SPN in the presence or absence of 0.50 mM supplemental NADPH in addition to 10 ng/ml enzyme and incubated for 1 h at 30°C. After incubation, samples were filtered using 50-kDa-cutoff filters (Amicon) to remove enzymes. Flowthrough was collected, diluted 1:1 with a mixture of H2O-AcN (95:5), 0.2 M formic acid (CH2O2), and 0.4 mM methionine sulfone (MeSO4), and stored at −80°C until analysis. Dilute salmon protein hydrolysate without enzymes was prepared in a similar fashion and used as a control. To quantify the TMA and TMAO in the salmon protein hydrolysate samples, another two series of calibration samples were prepared by spiking different quantities of TMA (50, 100, and 200 mg/liter) and TMAO (5, 10, 25, and 50 mg/liter).
CE-ESI-TOF-MS target analysis.
The data were obtained by CE (7100 Agilent) coupled to TOF (6224 Agilent). The separations occurred in a fused-silica capillary (total length, 100 cm, and inner diameter, 50 μm; Agilent). All separations were performed in normal polarity with a background electrolyte containing 1.0 M formic acid (FA) in 10% methanol (MeOH) (vol/vol) at 20°C. New capillaries were preconditioned with a flush of 1.0 M NaOH for 30 min, followed by MilliQ water for 30 min and the background electrolyte for 30 min. Before each analysis, the capillary was conditioned with a 5-min flush of the background electrolyte. The sheath liquid (0.6 ml/min) was MeOH-H2O (1:1) containing 1.0 mM FA with one reference mass of 121.0509 (purine, detected m/z [C5H4N4 + H]+), which allowed for correction and higher mass accuracy in the MS. The samples were hydrodynamically injected at 5,000 Pa for 17 s. Stacking was performed by applying the background electrolyte at 10,000 Pa for 10 s. The separation voltage was 30 kV, the internal pressure was 2,500 Pa, and the analyses were performed during 25 min. The MS parameters were as follows: fragmentor voltage, 125 V; skimmer voltage, 65 V; octopole voltage, 750 V; drying gas temperature, 200°C; flow rate, 10 liters/min; and capillary voltage, 3,500 V. The data were acquired in positive mode with a full scan from m/z 50 to 500 at a rate of 1.00 scan per second. Samples were analyzed in randomized runs. The analytical run was set up starting with the analysis of 5 injections of a pool of samples to equilibrate the system, followed by injections of the calibration samples and then the samples in a randomized order. After the final samples were injected, the series of calibration samples were injected again. The enzymatic reactions and salmon protein hydrolysate samples were analyzed in two independent analytical runs.
The corresponding peak areas were integrated using MassHunter Quantitative Analysis (B.09.00; Agilent). The final concentration per sample was calculated based on the peak area for the corresponding standard in a calibration curve. The linearity of the relative response versus concentration was previously assessed under the same analytical conditions (for enzymatic reactions, r2 = 0.9991 for TMA and r2 = 0.9944 for TMAO, and for salmon protein hydrolysate samples, r2 = 0.9973 for TMA and r2 = 0.9975 for TMAO).
ACKNOWLEDGMENTS
We thank Bjørn Liaset and the Biomega Group for providing us with salmon protein hydrolysate. We thank Ruth Matesanz at the CSIC-CIB for assistance during the CD analyses.
Funding from the Research Council of Norway (RCN grant number 280737) to G.E.K.B. is gratefully acknowledged. C.B. and D.R. acknowledge funding from the Spanish Ministry of Science, Innovation and Universities cofunded by FEDER (grant number RTI2018-095166-B-I00).
G.E.K.B. conceived the idea, and G.E.K.B., M.G., and M.F. planned, designed, and evaluated the experiments. M.G. analyzed all data. M.G. cloned, expressed, and purified all constructs and conducted biochemical analyses. P.P. and G.E.K.B. conducted SSN analysis, and P.P. performed bioinformatics. D.B. and C.B. planned, performed, and analyzed CE-ESI-TOF-MS experiments. J.C. cloned all constructs, and A.G.-M. and Ø.L. performed the initial cloning and expression testing. D.A. conducted CD experiments. M.G., P.P., and G.E.K.B. wrote the paper with input from M.F. All authors commented on the manuscript.
The authors declare no conflict of interests.