Aromatic compounds were purchased from Sigma-Aldrich (St Louis, MO, United States) and Acros Organics (Morris Plains, NJ, United States). Catalase, FAD, FMN, arabinose, antibiotics, Terrific broth (TB) and LB broth (Miller) (LB) were from Sigma-Aldrich (St Louis, MO, United States). Pefabloc SC and DNase I were obtained from Roche Diagnostics GmbH (Mannheim, Germany). Restriction enzymes and Pfu DNA polymerase were from Thermo Fischer Scientific (United States). 4-androstene-3,17-dione was from Merck (Oss, Netherlands). Crystallization kits were purchased from New Hampton (Aliso Viejo, CA, United States). Immobilized metal affinity chromatography columns (His GraviTrap) were from GE Healthcare Bioscience AB (Uppsala, Sweden). All other chemicals were from commercial sources and of the purest grade available.

All bacterial strains and primers used in this study are listed in Tables 1, 2.

The E. coli-Rhodococcus shuttle vector pRESQ (van der Geize et al., 2002) was modified by insertion of the RP4 oriT of pK18mobsacB (Schäfer et al., 1994) enabling trans-conjugal transfer of the resulting vector. For this, the oriT-region was amplified from pK18mobsacB by PCR using forward primer oriT-F and reverse primer oriT-R (Table 1). The obtained 549 bp PCR-product was cloned into the SmaI-site of pRESQ, resulting in pQmob. A duplicate region of 424 bp on pQmob was removed by deleting the 760 bp XbaI-BspHI fragment, yielding pQmobΔd. The egfp gene from pIJ8630 (Sun et al., 1999) was amplified by PCR using forward primer egfp-PciI-F, containing a PciI restriction site, and reverse primer egfp-PciI-R, also containing a PciI restriction site (Table 1). The 744 bp PciI-PciI fragment containing the egfp gene was cloned into the PciI-site of pQmobΔd to generate peGFPQ.

The R. jostii strain RHA1 genomic region consisting of gene ro00440, its promoter region and the prmA promoter (PprmA) (here referred to as region reg1-PprmA; GenBank accession number CP000431: nt 521345 - nt 523358) was amplified from genomic DNA of R. jostii RHA1 by PCR using forward primer reg1-BglII-F, containing a BglII restriction site, and reverse primer prmA-NdeI-R, containing an NdeI restriction site (Table 1). The 2014 bp BglII-NdeI reg1-PprmA fragment was cloned into the BglII-NdeI sites of peGFPQ, yielding prMOeGFPQ1.

The R. jostii RHA1 gene ro00452 and its promoter region (here referred to as region reg2; CP000431: nt 534363 – nt 536227) were amplified by PCR using forward primer reg2-PciI-F, containing a PciI restriction site and reverse primer reg2-PciI-R, also containing a PciI restriction site (Table 1). The 1880 bp PciI-PciI reg2 fragment was cloned into the PciI-site of prMOeGFPQ1, resulting in prMOeGFPQ2.

For construction of expression vector Q2+, the egfp gene of prMOeGFPQ2 was replaced with a multiple cloning site (MCS). For this, part of the MCS of pBluescript KS was amplified by PCR using forward primer MCS-NdeI-F, containing an NdeI restriction site and reverse primer MCS-PciI-R, containing a PciI restriction site (Table 1). The 125 bp NdeI-PciI MCS fragment was cloned into the NdeI–PciI site of prMOeGFPQ2, replacing the NdeI–PciI region containing egfp, resulting in the expression vector Q2+.

The 1321 bp NdeI-XmnI fragment of pBAD-3HB6H-His₆ (Montersino and van Berkel, 2012) containing the 3HB6H gene including the His₆-tag, was cloned into the NdeI-HindIII (Klenow-fragment treated) site of expression vector Q2+ to generate Q2+-3HB6H-His₆.

Rhodococcus cells were electroporated as described previously (van der Geize et al., 2000). Prior to electroporation, plasmid DNA was desalted by dialyzing 10 μL plasmid DNA for 30 min on a Millipore “V” Series filter disk (0.025 μm) floating on MiliQ water.

Cultures of R. jostii RHA1#2 were grown in LB broth supplemented with 50 μg⋅mL^-1 kanamycin at 30°C at 200 rpm. R. jostii RHA1#2 cells harboring Q2+3HB6H-His₆ were grown overnight in 3 mL LB broth, diluted 1:300 in 300 mL LB broth in a 3 L Erlenmeyer flask and grown for 20–24 h. Cultures were induced by adding 2 mM 4-androstene-3,17-dione dissolved in acetone. Growth was continued for 48 h after induction. Cells were harvested by centrifugation at 4°C and pellets were washed once with ice-cold 20 mM potassium phosphate, pH 7.2, containing 300 mM NaCl. After centrifugation at 4°C, cells were stored at -20°C.

The xlnD gene sequence encoding for Pa3HB6H (UniProt: Q9F131) was synthesized and subcloned in a pBAD vector by GeneArt (Invitrogen, Carlsbad, CA, United States). The resulting construct (pBAD-Pa3HB6H-His₆) was verified by automated sequencing of both strands and electroporated into E. coli TOP10 cells for recombinant expression.

For enzyme production, E. coli TOP10 cells, harboring the pBAD-Pa3HB6H-His₆ plasmid, were grown in TB medium at 37°C supplemented with 100 μg⋅mL^-1 ampicillin until an optical density (OD_{600 nm}) of 0.8 was reached. Expression was induced by the addition of 0.02% (w/v) arabinose and incubation was continued for 40 h at 17°C. Cells were harvested by centrifugation at 4°C and stored at -20°C.

Rj3HB6H was purified to apparent homogeneity using an Åkta Explorer chromatography system (GE-Healthcare). R. jostii RHA1#2 cells containing the recombinant protein were suspended in ice-cold 20 mM potassium phosphate, pH 7.2, containing 300 mM NaCl, 1 mM Pefabloc SC, 1 mg DNAse and 100 μM MgCl₂, and subsequently passed twice through a precooled French Pressure cell (SLM Aminco, SLM Instruments, Urbana, IL, United States) at 16,000 psi. The resulting homogenate was centrifuged at 25,000 × g for 45 min at 4°C to remove cell debris, and the supernatant was applied onto a Ni-NTA agarose column (16 mm × 50 mm) equilibrated with 20 mM potassium phosphate, pH 7.2, containing 300 mM NaCl. After washing with five volumes of equilibration buffer, the enzyme was eluted with 300 mM imidazole in equilibration buffer. The resulting Rj3HB6H fraction was supplemented with 100 μM FAD and loaded onto a Source Q-15 anion exchange column (16 mm × 90 mm), pre-equilibrated with 50 mM Bis-Tris, 0.1 mM EDTA, pH 7.2. After washing with two volumes of equilibrating buffer, the enzyme was eluted with a linear gradient of 0–1 M NaCl in the same buffer. Active fractions were pooled, concentrated to 10 mg⋅mL^-1 using ultrafiltration (Amicon 30 kDa cutoff filter), and applied onto a Superdex S-200 (26 mm × 600 mm) column running in 50 mM potassium phosphate, 150 mM NaCl, pH 7.2. Active fractions were concentrated to 10 mg⋅mL^-1 using ultrafiltration (Amicon 30 kDa cutoff filter) and dialyzed at 4°C against 50 mM Bis-Tris, pH 7.2. The final Rj3HB6H preparation showed a single band after SDS-PAGE. The specific activity of the purified enzyme was 21 U mg^-1 using the standard activity assay (Table 3A).

Pa3HB6H was purified to apparent homogeneity, applying essentially the same procedure as described above for Rj3HB6H. The final Pa3HB6H preparation showed a single band after SDS-PAGE. The specific activity of the purified enzyme was 34 U⋅mg^-1 using the standard activity assay (Table 3B).

Purified enzymes were flash frozen in 1 mL aliquots in liquid nitrogen and stored at -80°C. Before use, thawed enzyme samples were incubated with 0.1 mM FAD and excess FAD was removed using a gel filtration column (10 mm × 100 mm) containing Bio-Gel P-6DG.

Molar absorption coefficients of protein-bound FAD were determined from absorption spectra of Rj3HB6H and Pa3HB6H recorded in the presence and absence of 0.1% (w/v) SDS, assuming a molar absorption coefficient for free FAD of 11.3 mM^-1⋅cm^-1 at 450 nm. The enzyme concentration of Rj3HB6H was determined by measuring the absorbance at 453 nm using a molar absorption coefficient for protein-bound FAD of 10.3 mM^-1⋅cm^-1. The enzyme concentration of Pa3HB6H was determined by measuring the absorbance at 450 nm using a molar absorption coefficient for protein-bound FAD of 11.0 mM^-1⋅cm^-1. Rj3HB6H and Pa3HB6H activity was determined at 25°C by measuring NADH consumption at 360 nm (Montersino and van Berkel, 2012). The standard assay mixture contained 50 mM Tris-SO₄, pH 8.0, 200 μM 4-hydroxybenzoate and 250 μM NADH. Steady-state kinetic parameters were determined from measurements at 25°C in 50 mM Tris-SO₄, pH 8.0. Hydroxylation efficiencies were determined by oxygen consumption experiments, essentially as described before (Montersino and van Berkel, 2012).

Crystals of Rj3HB6H for structure determination were obtained by the sitting drop vapor diffusion method at 20°C by mixing equal volumes (2 μL) of protein and reservoir solutions. Protein solutions consisted of 30 mg⋅mL^-1 enzyme in 1 mM FAD, 2 mM 3-hydroxybenzoate, and 50 mM Bis-Tris, pH 7.2, whereas precipitant solutions consisted of 30% PEG 4000, 0.2 M lithium sulfate, and 0.1 M Tris-HCl, pH 8.5. Yellow crystals grew in 1 day.

X-ray diffraction data were collected at Grenoble and processed with the CCP4 package (Winn et al., 2011). The Rj3HB6H structure was solved by molecular replacement using the structure of a monomer of 3HB6H (pdb entry: 4BJZ) as search model. Crystallographic computing and model analysis were performed with COOT (Emsley et al., 2010), PHENIX (Adams et al., 2010) and the CCP4 package (Potterton et al., 2004). Pictures were generated with Pymol (Schrodinger, 2015) and CCP4 (Potterton et al., 2004). Data collection parameters and refinement statistics are presented in Table 4.

The atomic coordinates and structure factors of Rj3HB6H (code 5HYM) have been deposited in the Protein Data Bank¹.

Extraction and identification of protein-bound lipids from Rj3HB6H and Pa3HB6H was performed as described for 3HB6H (Montersino et al., 2013). For nanoflow ESI-MS analysis under native conditions, enzyme samples were prepared in 50 mM ammonium acetate, pH 6.8. For analysis under denaturing conditions, enzyme samples were diluted either in 50% acetonitrile with 0.2% formic acid or in 5% formic acid. Native MS analysis was performed using a LC-T nanoflow ESI orthogonal TOF mass spectrometer (Micromass, Manchester, United Kingdom) in positive ion mode with a capillary voltage of 1.3 kV. The cone voltage was varied between 90 and 150 V and source pressure was set to 6.9 mbar to enhance transmission of large ions. Lipid identification was performed using a Quattro Ultima nanoflow triple quadrupole mass spectrometer (Micromass, Manchester, United Kingdom) in negative ion mode, with a capillary voltage of 1.3 kV and a cone voltage of 150 V. For MS/MS analysis, argon was supplied in the collision cell (2.0 × 10^-3 bar). Collision energy was adjusted to gain optimal fragmentation. Both mass spectrometers were equipped with a Z-spray nano-electrospray ionization source. Measurements were performed by using gold-coated needles, made from borosilicate glass capillaries (Kwik-Fill; World precision Instruments, Sarasota) on a P-97 puller (from Sutter Instruments, Novato, CA, United States). Needles were coated with a gold layer using an Edwards Scancoat six Pirani 501 sputter coater (Edwards laboratories, Milpitas, CA, United States). All TOF spectra were mass calibrated by using an aqueous solution of cesium iodide (25 mg⋅mL^-1).

Protein sequences were retrieved using protein resources from the National Centre for Biotechnology Information² and UniProt Database³. Multiple sequence alignments were made using CLUSTALW (Thompson et al., 1994). Phylogenetic plots were made using FigTree⁴.

Expression of the 3HB6H gene from R. jostii RHA1#2 yielded about 7 mg of purified Rj3HB6H protein from 10 g wet cells (Table 3A). Rj3HB6H displayed the same absorption spectrum as 3HB6H, with maxima at 274, 383, and 453 nm and a shoulder at 480 nm (Montersino and van Berkel, 2012). A molar absorption coefficient of protein-bound flavin, 𝜀₄₅₃ = 10.3 mM^-1 cm^-1, was used for both proteins.

Determination of steady-state kinetic parameters revealed that Rj3HB6H behaves similarly as 3HB6H using 3-hydroxybenzoate as variable substrate and fixed NADH concentration (k_cat = (20 ± 1) s^-1; K_M = (35 ± 3) μM; k_cat/K_M = (5.7 ± 0.8) × 10⁵ s^-1⋅M^-1) and with variable concentration of NADH (preferred coenzyme) and fixed 3-hydroxybenzoate concentration (k_cat = (20 ± 1) s^-1; K_M = (68 ± 5) μM; k_cat/K_M = (3.0 ± 0.4) × 10⁵ s^-1⋅M^-1). Rj3HB6H displays a very low NADH oxidase activity (<1 U⋅mg^-1). Uncoupling of hydroxylation of 3-hydroxybenzoate occurs to a minor extent (less than 10%), while 2,5-dihydroxybenzoate is a strong non-substrate effector (k_cat = (6 ± 0.8) s^-1; K_M = (150 ± 30) μM; k_cat/K_M = (4.0 ± 1.3) × 10⁴ s^-1⋅M^-1), efficiently stimulating the rate of flavin reduction by NADH (Montersino and van Berkel, 2012; Sucharitakul et al., 2012, 2013; Ni et al., 2016).

Rj3HB6H crystals grew in similar conditions as found for 3HB6H, and are isomorphous to those of 3HB6H, where lithium sulfate was present instead of sodium acetate. The three-dimensional structure of Rj3HB6H was solved at 2.3 Å resolution by molecular replacement (Table 4). The isoalloxazine moiety of FAD was refined with full occupancy in the in conformation. Similar to the crystallographic analysis of 3HB6H, no substrate could be detected in the active site of the enzyme, despite presence of excess 3-hydroxybenzoate in the crystallization drop. The protein crystallizes as a dimer, just as 3HB6H (Montersino et al., 2013), and contains a phospholipid molecule in each subunit. The electron density of the phospholipid in the crystal structure was refined as two acyl chains, one of twelve and one of seventeen carbon units. Superimposition of the Rj3HB6H and 3HB6H models (root mean square deviation = 0.22 Å) shows minor deviations (Figure 2A). The phospholipid is located in a tunnel, which runs from the dimer interface to the active site (Figure 2B), and interacts with the opposite monomer. The phosphate group resides at the protein surface near Arg350 and Lys385, and the electron density of the headgroup is consistent with the presence of a cyclohexanehexol moiety (Figure 3A).

Assignment of protein-bound phospholipids was achieved by ESI-MS analysis of the low molecular weight components extracted from denatured Rj3HB6H. The mass spectrum in negative mode (Figure 3B) displayed three main peaks with m/z values of 823, 837, and 851. From the MS pattern it was evident that Rj3HB6H binds a phospholipid with a bigger headgroup compared to that of the lipid found in 3HB6H.

Fragmentation analysis and comparison of data to reference lipid MS spectra led to a match of the obtained mass peaks with those of PI, having aliphatic chains containing 15 to 19 carbons. Peak 1 (m/z 823) is assigned to PI (15:0/18:0), peak 2 (m/z 837) is assigned to PI (16:0/18:0) (Sharp et al., 2007; Morita et al., 2011) and peak 3 ((m/z 851) is assigned to PI (16:0/19:0) with alternate acylate form (tuberculostearic acid) (Drage et al., 2010).

Typical fragmentation of PI was visible in the MS/MS spectra by signature peaks with m/z values of 153, 223, 241, and 297, representing glycerol phosphate water (m/z 153) and inositol headgroup fragments (Lang and Philp, 1998; Pulfer and Murphy, 2003; Oursel et al., 2007) (data not shown). Minor peaks at approximately 2 m/z values lower than the identified peaks represent the same PI, containing one unsaturated bond.

To gain further insight into the enzyme–lipid interaction, we determined the oligomeric protein composition of 3HB6H and Rj3HB6H using native MS (Leney and Heck, 2017). As a first step, we determined the experimental masses of the denatured proteins. The measured values (46,766 ± 4 Da for 3HB6H and 46,761 ± 2 Da for Rj3HB6H) agree with the mass deduced from the primary sequence, lacking the N-terminal methionine.

Native MS of 3HB6H showed eight charge states corresponding to five different protein forms (Figure 4A and Table 5). The charge state distribution ions +12, +13, and +14 represent the monomeric apoprotein (average mass 46,835 ± 5 Da; red stars), the monomeric holoprotein (average mass 47603 ± 6 Da; green stars), and the monomeric holoprotein containing additionally one PG/PE molecule (average mass 48,312 ± 7 Da; blue stars). The charge state distribution ions +18, +19, +20, +21, and +22 predominantly represent the dimeric holoprotein with either one or two PG/PE molecules bound (average mass 95,868 ± 16 Da; orange stars, and 96,643 ± 14 Da; purple stars, respectively). Tandem MS experiments revealed that one, two, three, or four ligands can be expelled from 3HB6H. The assignment of bound ligands was made on the basis of total mass increase and comparison with the mass of the native apoprotein (Table 5).

Native MS of Rj3HB6H also showed a range of charge state distributions (Figure 4B and Table 5). The charge state distribution ions +12, +13, and +14 represent the monomeric apoprotein (average mass 46,829 ± 1 Da; red star), the monomeric holoprotein (average mass 47613 ± 1 Da; green stars), and the monomeric holoprotein containing one PI molecule (average mass 48,458 ± 2 Da; blue star). The latter species differs from the related 3HB6H species (Figure 4A, blue stars) because it has a bigger lipid headgroup. The major species in the native mass spectrum of Rj3HB6H corresponds to the holo-Rj3HB6H dimer with PI bound to both subunits (average mass 96,938 ± 32 Da; Figure 4B, yellow stars). Only by magnification it is possible to detect a minor peak representing the holo-Rj3HB6H dimer with one PI bound (average mass 96,128 ± 5 Da; Figure 4B, orange stars). A cartoon of the different subunit compositions of 3HB6H is presented in Figure 4D.

To analyze whether the lipid-binding site of 3HB6H is conserved among species, we explored the natural diversity of 3HB6H enzymes. 3HB6H activity has been reported for Gram-positive and Gram-negative bacteria and for yeasts. Besides from the R. jostii prototype, the enzymes from Klebsiella pneumonia M5a1 (Suárez et al., 1995; Liu et al., 2005), Pseudomonas alcaligenes NCIMB 9867 (Gao et al., 2005), Polaromonas naphthalenivorans CJ2 (Park et al., 2007), Corynebacterium glutamicum ATCC 12032 (Yang et al., 2010), Rhodococcus sp. NCIMB 12038 (Liu et al., 2011) and Candida parapsilosis (Holesova et al., 2011) have been characterized to some extent.

From the structural data of the R. jostii 3HB6H enzyme and the multiple sequence alignment presented in Figure 5 it can be inferred that residues directly involved in lipid binding in Rj3HB6H are not always conserved in the orthologs; among the bacterial enzymes studied, most sequence divergence occurs in 3HB6H from P. alcaligenes NCIMB 9867 (Pa3HB6H). This prompted us to study the lipid binding properties of the Pseudomonas 3HB6H enzyme.

Expression of the Pa3HB6H gene in E. coli TOP10 cells yielded about 10 mg of enzyme from a 1 L batch culture. Purified Pa3HB6H had a specific activity of 34 U⋅mg^-1 (Table 3B) and migrated in SDS-PAGE as a single band with an apparent subunit mass of 47 kDa (not shown). ESI-MS established that native Pa3HB6H is a dimer, and not a trimer as suggested earlier (Gao et al., 2005), and that the enzyme indeed contains lipids (Figure 4C and Table 5). The mass spectrum of extracted lipids showed peaks with m/z values characteristic of PG and PE with aliphatic chains ranging from 14 to 19 carbons, similar to the previously identified lipids in 3HB6H from R. jostii RHA1 produced in E. coli (Montersino et al., 2013).