The 1335-bp full-length tmm gene from Pelagibacter sp. HTCC7211 was synthesized by the Beijing Genomics Institute (China). The gene was then subcloned into the pET28a (Novagen, United States) vector with an N-terminal His tag. The point mutations in Tmm₇₂₁₁ were introduced using PCR-based method and verified by DNA sequencing. The Tmm₇₂₁₁ protein and its mutants were expressed in E. coli BL21 (DE3). The cells were cultured at 37°C in Lysogeny Broth medium to an OD₆₀₀ of 0.8–1.0 and then induced at 20°C for 14 h with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). The proteins were purified first with Ni²⁺-NTA resin (Qiagen, Germany) and then fractionated on a Superdex-200 column (GE Healthcare, United States). The protein concentration was determined with the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, United States), and a nine-point calibration curve of bovine serum albumin (BSA) standards was used according to the user guide.

A Superose 6 column was used for gel filtration analysis, because it possesses a wider fractionation range than the Superdex-200 column. The Superose 6 column was calibrated in the buffer containing 10 mM Tris-HCl (pH 8.0) and 100 mM NaCl using the following standards from GE Healthcare: thyroglobulin (669 kDa), ferritin (440 kDa), aldolase (158 kDa), conalbumin (75 kDa), carbonic anhydrase (29 kDa), ribonuclease A (13.7 kDa), and aprotinin (6.5 kDa). The void volume of Superose 6 column was determined with Blue Dextran 2000 (2,000 kDa).

The UV spectra of Tmm₇₂₁₁ (0.1 mM protein in the buffer containing 10 mM Tris-HCl (pH 8.0) and 100 mM NaCl) were measured by a V550 UV/VIS spectrophotometer (Jasco, Japan) in a cell with 1.0 cm path length (Response: Medium; Band width: 1.0 nm). The spectra of the mixture of Tmm₇₂₁₁ and NADPH were measured immediately after NADPH (0.1 mM) was added into the protein solution.

The DMSO produced by the enzymatic activity of Tmm₇₂₁₁ toward DMS was measured by high performance liquid chromatography (HPLC) (Dionex, America) on a SunFire C₁₈ column (Waters, America). The detection wavelength was 210 nm because DMSO exhibited an absorbance maximum at ∼210 nm. The samples were eluted in HPLC buffer (2.5% (v/v) acetonitrile, 0.2% (v/v) phosphoric acid in double-distilled H₂O) over 20 min at a flow rate of 1 ml/min. The reaction system contained 6 mM DMS (Sigma-Aldrich, America), 1.5 mM NADPH (Sigma-Aldrich, America), 0.15 mM Tmm₇₂₁₁, 10 mM Tris-HCl (pH 7.0) and 100 mM NaCl. The reaction was performed at 25°C, pH 7.0 for 3 h, and terminated by adding 10% phosphoric acid. The reaction system was centrifuged at 15,000 g for 15 min, and then the supernatant (20 μl) was injected for HPLC analysis. The control group had the same reaction system except that Tmm₇₂₁₁ was not added.

In the absence of DMS, the consumption of NADPH was less than 3% of that in the presence of DMS, indicating the NADPH-oxidase activity (also known as uncoupling) of Tmm₇₂₁₁ is rather weak under the experimental conditions. Because monitoring NADPH oxidation is more sensitive than monitoring DMSO formation, the enzymatic activity of Tmm₇₂₁₁ was measured by following the decrease of absorbance at 340 nm (ε₃₄₀ = 6.22 mM^–1 cm^–1 for NADPH) (Alfieri et al., 2008). The reaction mixture for detecting the enzymatic activity of Tmm₇₂₁₁ contains 1 μM Tmm₇₂₁₁, 0.25 mM NADPH, 1 mM DMS, 10 mM Tris-HCl (pH 7.0) and 100 mM NaCl. The reaction mixture without Tmm₇₂₁₁ was set as the control. For the measurements of the apparent K_M values of Tmm₇₂₁₁, substrate (DMS, TMA, DMA or methimazole) of different concentrations was added into the reaction system containing 1 μM Tmm₇₂₁₁ and 0.25 mM NADPH. For the measurements of the apparent K_M values of Tmm₇₂₁₁ and its mutants toward NADPH, different concentrations of NADPH were added into the reaction system containing 1.5 μM purified enzyme and 1 mM DMS. The optimal pH and the optimal temperature of Tmm₇₂₁₁ were determined using DMS as the substrate. For measurement of the optimal temperature of Tmm₇₂₁₁, a buffer containing 10 mM Tri-HCl (pH 8.0) and 100 mM NaCl was pre-incubated at different temperatures for 30 min, and then 1 μM Tmm₇₂₁₁, 0.25 mM NADPH and 1 mM DMS were added into the buffer. The mixture was incubated at different temperatures for 6 min before detection of NADPH oxidation at 340 nm using a V550 UV/VIS spectrophotometer (Jasco, Japan). The optimum of pH was examined at 25°C (the optimal temperature for Tmm₇₂₁₁ enzymatic activity) using Bis-Tris buffer for pH 6–7, Tris buffer for pH 7–9 and glycine buffer for pH 9–10.

The purified Tmm₇₂₁₁ protein was concentrated to ∼8 mg ml^–1 in 10 mM Tris-HCl (pH 8.0) and 100 mM NaCl. To obtain crystals of Tmm₇₂₁₁, NADPH with a final concentration of 5 mM was added into the protein solution before crystallization. Initial crystallization trials for Tmm₇₂₁₁ were performed at 18°C using the sitting drop vapor diffusion method. Diffraction-quality crystals of Tmm₇₂₁₁ were obtained in hanging drops containing 0.2 M ammonium citrate tribasic, 0.1 M imidazole (pH 7.0) and 20% (w/v) polyethylene glycol monomethyl ether 2,000 at 18°C after a 3-week incubation. To obtain the crystals of Tmm₇₂₁₁ soaked with DMS, Tmm₇₂₁₁ crystals were soaked in 20 mM DMS for 5 and 20 min, respectively. X-ray diffraction data were collected on the BL17U1 (Wang et al., 2018) and BL18U1 beamlines at the Shanghai Synchrotron Radiation Facility. The initial diffraction data sets were processed using the HKL3000 program with its default settings (Minor et al., 2006).

The crystals of Tmm₇₂₁₁ and Tmm₇₂₁₁ soaked with DMS belong to the P2₁ space group. The crystal structures of Tmm₇₂₁₁ and Tmm₇₂₁₁ soaked with DMS were determined by molecular replacement using the CCP4 program phaser (Winn et al., 2011) with the crystal structure of a bacterial Tmm (PDB code: 5IPY) as the search model. The refinement of these structures were performed using WinCoot (Emsley et al., 2010) and Phenix (Adams et al., 2010). Default parameters in CCP4, WinCoot and Phenix were used. All the structure figures were processed using the program PyMOL.¹

CD spectroscopic assays for Tmm₇₂₁₁ and all its mutants were carried out on a J-1,500 Spectrometer (Jasco, Japan) in a 1 mm pathlength cuvette at 25°C. The concentration of the proteins was 8.0 μM in the buffer of 10 mM Tris-HCl (pH 8.0) containing 100 mM NaCl. The buffer without proteins was used for baseline and blank measurements. The spectra were collected from 250 to 200 nm at a scan speed of 500 nm min^–1 with a band width of 1 nm. Each sample was scanned for three times. The noise level is < 0.05 mdeg.

Related protein sequences DddD (Pseudomonas putida, WP_062573753.1), DddK (Candidatus Pelagibacter ubique HTCC1062, WP_011281678.1), DddP (Mesorhizobium, WP_109668646.1), DddQ (Mesorhizobium loti, WP_10966 8666.1), DddW (Ruegeria pomeroyi, WP_011046214.1), DddL (Puniceibacterium antarcticum SM1211, WP_099909581.1), DddY (Alcaligenes faecalis, WP_123051132.1), DMSOR (Rhodobacter capsulatus, Q52675.2), Tmm (Pelagibaca abyssi, APZ51459.1), DdhA (Sagittula stellata E-37, EBA07058.1), MddA (Pseudomonas deceptionensis, WP_048359798.1), DsoB (Acinetobacter sp. 20B, BAA23331.1) and DmoA (Hyphomicrobium sulfonivorans, E9JFX9.1) were obtained from National Center for Biotechnology Information (NCBI) database² as seed sequences. For multifunctional strains screening, the seed sequences were used to search against the genomes of isolated strains on the IMG/M metagenomics database (Chen et al., 2019) with parameters of similarity > 40%, E-value of < 10^–50 and coverage > 70% to elevate the accuracy and precision of blast hits. Data processing was performed via scripts compiled in Python code.³ The biological networks of related proteins were built via software Cytoscape 3.8.0 (Kohl et al., 2011).

The structures of Tmm₇₂₁₁, Tmm₇₂₁₁-5-min and Tmm₇₂₁₁-20-min have been deposited in the Protein Data Bank (PDB) under the accession codes 7D4K, 7D4M, and 7D4N, respectively.

The tmm gene of Pelagibacter sp. HTCC7211 contains 1335 nucleotides and encodes a protein of 444 amino acid residues, with a calculated molecular mass of 52 kDa. Tmm₇₂₁₁ shares ∼53% amino acid sequence identity with RnTmm, a previously reported Tmm homolog from an MRC strain Roseovarius nubinhibens ISM (Li et al., 2017). Full-length tmm of strain HTCC7211 was synthesized and was expressed in E. coli BL21 (DE3) cells, and the recombinant Tmm₇₂₁₁ was purified (Figure 1A) and characterized. The purified Tmm₇₂₁₁ is yellow, suggesting that FAD has been already bound in the recombinant Tmm₇₂₁₁ during protein expression in E. coli, which is further supported by spectroscopic analysis. The purified Tmm₇₂₁₁ exhibited the typical absorbance maxima (around 372 and 442 nm) of fully oxidized FMOs (Figure 1B; Alfieri et al., 2008; Orru et al., 2010). Addition of equimolar amount of NADPH should lead to the formation of the enzyme-(hydro)peroxyflavin-NADP⁺ complex exhibiting a typical absorbance maximum at around 360 nm (Alfieri et al., 2008; Orru et al., 2010). However, the absorption spectrum showed that the absorbance maximum of Tmm₇₂₁₁ with the addition of equimolar amount of NADPH was around 350 nm (Figure 1B). Because NADPH absorbs at 340 nm (Alfieri et al., 2008), this spectrum probably reflected a mixture of the enzyme-(hydro)peroxyflavin-NADP⁺ complex and some residual NADPH, which may due to some inactive enzymes in the purified Tmm₇₂₁₁ solution. Incubation of recombinant Tmm₇₂₁₁ with DMS and NADPH yielded DMSO and NADP⁺ (Figure 1C), demonstrating that Tmm₇₂₁₁ has DMS oxidation activity in vitro. The optimal temperature for Tmm₇₂₁₁ enzymatic activity toward DMS was ∼25°C (Figure 1D), and the optimal pH was 7.0 (Figure 1E). The optimal temperature of Tmm₇₂₁₁ is lower than that of RnTmm toward TMA (30°C) (Li et al., 2017). Furthermore, Tmm₇₂₁₁ only retained ∼40% of its highest enzymatic activity at 30°C, whereas RnTmm still retained ∼70% of its highest enzymatic activity at 40°C (Li et al., 2017), suggesting that Tmm₇₂₁₁ is more sensitive to high temperature than RnTmm.

The substrate specificity of Tmm₇₂₁₁ was also analyzed. Tmm₇₂₁₁ can oxidize DMS, TMA, DMA, and methimazole, with TMA showing the highest affinity (Table 1). In general, the apparent K_M values of Tmm₇₂₁₁ to different substrates are slightly higher than those of RnTmm, and the k_cat values of Tmm₇₂₁₁ are lower (Table 1; Li et al., 2017), indicating that the enzymatic activity of Tmm₇₂₁₁ is lower than that of RnTmm in vitro.

To gain insight into the putative active site of Tmm₇₂₁₁, we solved the crystal structure of Tmm₇₂₁₁ to 1.8 Å (Table 2). The crystals of Tmm₇₂₁₁ belong to the P2₁ space group, with two molecules arranged as a dimer in an asymmetric unit (Figures 2A,B). Gel filtration analysis (Figure 2C) indicated that Tmm₇₂₁₁ functions as a dimer in solution, which is supported by the result of the PISA server prediction.⁴ After structural refinement, the NADP⁺ and FAD molecules can be clearly observed in the structure (Figure 2A). The overall structure of Tmm₇₂₁₁ is similar to those of other reported bacterial FMOs (Alfieri et al., 2008; Cho et al., 2011; Li et al., 2017), with the root mean square deviations (RMSDs) between Tmm₇₂₁₁ and other bacterial FMOs of no more than 0.6 Å. Tmm₇₂₁₁ also comprises an NADPH binding domain and an FAD binding domain (Figure 2B). These two domains are connected through two hinge regions (Ser163–Pro168 and Cys268–Leu272) (Figure 2B).

To obtain the crystal structure of Tmm₇₂₁₁ in complex with DMS, we first tried to co-crystalize Tmm₇₂₁₁ with DMS. However, this failed, probably due to the volatile nature of DMS that has a low boiling point (∼37°C) in the crystallization buffer. Next, we tried the soaking method and solved two crystal structures of Tmm₇₂₁₁ soaked with DMS for different soaking time (Table 2). For briefness, the crystal structures of Tmm₇₂₁₁ soaked with DMS for 5 min and for 20 min were termed as Tmm₇₂₁₁-5-min and Tmm₇₂₁₁-20-min, respectively. The overall structures of Tmm₇₂₁₁ soaked with DMS are similar to that of Tmm₇₂₁₁, with the RMSD between Tmm₇₂₁₁ and Tmm₇₂₁₁-5-min of 0.1 Å, and the RMSD between Tmm₇₂₁₁ and Tmm₇₂₁₁-20-min of 0.2 Å.

From the surface view of Tmm₇₂₁₁, we can only observe part of the NADP⁺ molecule and the FAD molecule was not visible (Figure 3A). The nicotinamide ring of NADP⁺ is located inside Tmm₇₂₁₁, and the entire FAD molecule is deeply bound in the protein (Figures 3A,B). The binding of NADP⁺ and FAD mainly depends on hydrogen bonds formed between Tmm₇₂₁₁ residues and them (Figures 3C,D). For NADP⁺ binding, residues Trp70 and Arg409 form hydrogen bonds with the nicotinamide ring, Asn72 and Gln315 interact with the ribose ring via water-mediated hydrogen-bonds, and Tyr170, Ser202, Ser203, Ser205, Arg226, His227, and Asn288 interact with the other parts of NADP⁺ (Figure 3C). For FAD binding, residues Asn72 and Thr318 interact with the isoalloxazine ring, Glu37 forms hydrogen bonds with the ribose ring, Val125 forms a hydrogen bond with the adenine moiety, and Gly10, Leu45, Trp46, Gly160, Ser163, and Gln315 interact with the other parts of FAD (Figure 3D).

To confirm the importance of Tmm₇₂₁₁ residues involved in binding NADP⁺ and FAD, we generated site-directed mutations to the related residues and quantified the enzymatic activities of the mutants. All mutants had significantly decreased activity (Figure 3E), suggesting that these residues play important roles for the correct binding of NADP⁺ or FAD. Moreover, mutants Asn72Ala, Ser203Ala, Arg226Ala, His227Ala, Asn288Ala, and Arg409Ala all exhibited higher apparent K_M values toward NADPH than wild type (WT) Tmm₇₂₁₁ (Table 3), further supporting their roles in binding NADP⁺/NADPH. CD spectroscopy analysis showed that the secondary structures of the mutants exhibited little deviation from that of WT Tmm₇₂₁₁ (Figure 3F), indicating that the decreases in the enzymatic activities of the mutants are caused by residue replacement rather than structural changes.

To elucidate the catalytic mechanism of Tmm₇₂₁₁ for DMS oxidation, it is important to ascertain the location of DMS. Despite the two structures of Tmm₇₂₁₁ soaked with DMS were solved, the explicit electron density of DMS in the structures could not be identified. Previous structural analyses demonstrated that the substrate of bacterial FMOs with a ring structure, such as indole and methimazole, is located in the position of the nicotinamide ring of NADP⁺, forming stacking interactions with the isoalloxazine ring of FAD (Eswaramoorthy et al., 2006; Cho et al., 2011; Li et al., 2017). There is no direct interaction between the residues of bacterial FMOs and the substrates (Eswaramoorthy et al., 2006; Cho et al., 2011; Li et al., 2017). For substrates with no ring structure, such as DMS and TMA, there may be no effective interactions to stabilize their conformations, which may be the reason why we could not find the DMS molecule in the structures of Tmm₇₂₁₁ soaked with DMS.

By comparing the structures of Tmm₇₂₁₁, Tmm₇₂₁₁-5-min and Tmm₇₂₁₁-20-min, we noticed that with the extension of soaking time, the electron densities of the nicotinamide ring and the ribose ring of NADP⁺ become increasingly weaker (Figures 4A–C), indicating that the nicotinamide ring and the ribose ring become flexible during soaking. Combined with previous studies that the indole and methimazole molecules located in the binding site of NADP⁺ (Eswaramoorthy et al., 2006; Cho et al., 2011; Li et al., 2017), this result suggests that DMS may also be bound in the position of the nicotinamide ring of NADP⁺, and that the entry of DMS may repel NADP⁺, leading to a conformational change of the nicotinamide ring and the ribose ring.

The conformational change of NADP⁺ was also observed in RnTmm when soaked with TMA, and this conformational change was shown to be important for TMA oxidation (Li et al., 2017). After conformational change, the ribose ring of NADP⁺ in RnTmm forms a hydrogen bond with Asp317, shutting off the substrate entrance to promote a protected micro-environment for catalysis (Li et al., 2017). Because the electron densities of the ribose ring in Tmm₇₂₁₁-5-min and Tmm₇₂₁₁-20-min are rather poor (Figures 4B,C), we could not ascertain whether the ribose ring can form a hydrogen bond with Asp314 of Tmm₇₂₁₁, the equivalent residue to Asp317 of RnTmm. To further probe this, we generated mutants Asp314Ala and Asp314Glu, and measured their enzymatic activities. The enzymatic activity and the apparent K_M of Tmm₇₂₁₁ toward NADPH are only slightly affected by Asp314Glu mutation (Figure 3E and Table 3), probably due to the similar properties of aspartic acid and glutamic acid. However, although the residue Asp314 is far away from the catalytic center of Tmm₇₂₁₁, the mutation of Asp314 to alanine decreased the activity of Tmm₇₂₁₁ significantly (Figure 3E), suggesting that Asp314 is involved in the catalytic reaction of Tmm₇₂₁₁. The CD spectrum of the mutant Asp314Ala was indistinguishable from that of WT Tmm₇₂₁₁, suggesting that the enzymatic activity loss in the mutant is caused by residue replacement rather than structural alteration of the enzyme (Figure 3F). In addition, the Asp314Ala mutation increased the apparent K_M of Tmm₇₂₁₁ toward NADPH (Figures 4D,E), suggesting that this residue participates in binding NADP⁺/NADPH. Because structural analysis shows that the residue Asp314 is too far away to participate in NADP⁺ binding before DMS enters the catalytic center (Figure 3C), this result suggests that, after DMS enters the catalytic center, NADP⁺ in Tmm₇₂₁₁ likely undergoes a conformational change and forms a new hydrogen bond with Asp314, which is important for the catalysis of DMS oxidation.