ORFs encoding the TfAcCCB (Tfu_2555; WP_011292978), TfAcCCA (Tfu_2557; WP_011292980), and TfAcCCE (Tfu_2556; WP_011292979) proteins were initially identified using The Thioester-active enzYme database (ThYme) (http://www.enzyme.cbirc.iastate.edu/) (Cantu et al., 2010, 2011; Chen et al., 2011). The respective ORFs were codon optimized for expression in E. coli and chemically synthesized as “gBlocks” (Integrated DNA Technologies, Coralville, IA, USA). The assembled ORFs were PCR amplified, using suitable primers (Supplementary Table 1), and the resulting assembled fragments were used to construct expression vectors. These vectors included pCDFDuet-1 (MilliporeSigma, MA, USA), which was used for the co-expression of N-terminal 6X-His tagged TfAcCCB (cloned into MCSI at the restriction enzyme sites, BamHI and AflII), with a C-terminal S-tagged TfAcCCE (cloned into MCSII at the NdeI and XhoI restriction enzyme sites) (Raines et al., 2000). The TfAcCCA ORF was cloned into pET30F vector, at the restriction enzymes sites, BamHI and XhoI. TfAcCCA and TfAcCCE were individually expressed from pETDuet-1 vectors, with the ORFs cloned between the NcoI and NotI restriction sites of the MCS1; in these constructs TfAcCCA and TfAcCCE were expressed without any tag. The authenticity of all vector constructs was determined by DNA sequencing, conducted at Iowa State University's DNA Facility.

Expression plasmid constructs were harbored in E. coli strain BL21 λ(DE3), which were selected by the ability of the strain to grow in LB medium supplemented with the appropriate antibiotic: streptomycin (50 μg/ml) for pCDFDuet-1-based vectors, ampicillin (50 μg/ml) and kanamycin (30 μg/ml) for pET30F-based vectors. Overnight liquid pre-cultures were diluted 1:100 into fresh media (500 ml), and growth was continued at 37°C with agitation at a rate of 250 revolutions/min. Protein expression was induced by adding 0.4 mM IPTG when the A600 of the culture reached ~0.6–0.8. Following another 8 h of growth, cells were harvested by centrifugation, and cell pellets were flash frozen in liquid N₂ and stored at −80°C until further use. Additional experiments were conducted to ensure that biotinylation was not limiting the recovery of active AcCCase enzyme. Specifically, cultures of strains expressing TfAcCCase components were supplemented with 1 mg/ml biotin. Comparative analysis of recovered enzyme from cultures grown with and without biotin indicted that biotinylation was not limiting the recovery of active AcCCA or AcCCase enzyme.

Cell pellets were suspended in a buffer consisting of 50 mM Tris-Cl, pH 8, 300 mM NaCl, 0.75 mM DTT, 1 mM EDTA, 10% glycerol and the recommended amount of Pierce EDTA-free protease inhibitor (Thermo Scientific, Waltham, MA). The cell suspension was sonicated on ice with 7–10 bursts, each of 15 s duration, followed by a minute of cooling. The sonicated lysate was clarified by centrifugation at 20,000 g for 30 min. The supernatant was filtered with a 0.45 μ filter (Corning Lifesciences, Corning, NY, USA) and applied to a 5 ml cOmplete™ His-tag purification Ni-NTA column (Roche Diagnostics Gmbh, Mannheim, Germany), which was equilibrated with Column Buffer (50 mM Tris-Cl, pH 8, 300 mM NaCl, 0.75 mM DTT, 1 mM EDTA, and 10% glycerol) containing 20 mM imidazole. Subsequently, the column was sequentially washed with Column Buffer containing 40 and 60 mM imidazole, and finally the His-tagged protein(s) were eluted with 10 bed volumes of Column Buffer containing 150 mM imidazole. Eluted protein-fractions were dialyzed overnight against 100 mM potassium phosphate buffer, pH 7.6, containing 0.75 mM DTT, 1 mM EDTA and 20% glycerol, and finally concentrated using an Amicon® Ultra-15 centrifugal filters (MilliporeSigma, Billerica, MA, USA). Protein samples were flash frozen in liquid N₂ and stored at −80°C. Purified protein fractions were analyzed by SDS-PAGE (Laemmli, 1970), and protein bands were visualized by staining the gels with Coomassie Brilliant Blue R-250. The relative intensity of the stained protein bands was quantified with ChemiDoc XRS⁺ using the software Image Lab 6.0 (Bio-Rad, Hercules, CA).

The D427I variant of TfAcCCB ORF was carried out using the Qiagen QuickChange kit (Agilent Technologies, Santa Clara, CA, USA), using the primer sequence listed in Supplementary Table 1. The mutation was confirmed by DNA sequencing (DNA Facility, Iowa State University).

The enzyme catalyzed carboxylation reaction was assayed via the coupled hydrolysis of ATP. Specifically, the rate of ADP formation was determined with the combination of pyruvate kinase and lactate dehydrogenase in the presence of excess phosphoenol pyruvate, resulting in the reduction in A340 due to the conversion of NADH to NAD⁺ (Grassl, 1974). Each reaction contained 100 mM potassium phosphate, pH 7.6, 5 mM MgCl₂, 3 mM ATP, 1 mM NADH, 0.3 mg/ml BSA, 50 mM NaHCO₃, 0.5 mM PEP, ~5 Units of pyruvate kinase and ~7 Units of lactate dehydrogenase. Reaction mixture was incubated at 37°C for 10 min, and the enzyme catalyzed reaction was initiated by the addition of the acyl-CoA substrate (MilliporeSigma, MA, USA). Enzyme catalysis was monitored continuously by reading absorbance at 340 nm using BioTek ELx808™ Absorbance Microplate Reader (BioTek US, Winooski, VT, USA).

The molecular weight of purified protein complexes was determined by FPLC using a Superdex 200 10/300 GL gel filtration column (~24 ml bed volume) (GE Healthcare Life Sciences, Pittsburg, PA) using an ÄKTA FPLC system (GE Healthcare Bio-Sciences). The column was equilibrated with 100 mM potassium phosphate buffer, pH 7.6, 0.75 mM DTT, 1 mM EDTA and 20% glycerol. The column was calibrated using BioRad gel filtration molecular weight standards, which contained: bovine thyroglobulin (670 kDa), bovine γ-globulin (158 kDa), chicken ovalbumin (44 kDa), horse myoglobin (17 kDa), and Vitamin B12 (1.45 kDa). The elution volumes of these standards were compared to those of different AcCCase preparations to determine the molecular weight of the latter.

The ThYme database (Cantu et al., 2010, 2011; Chen et al., 2011) was queried with S. coelicolor AcCCase sequences (WP_003973461.1; ScAcCCB, AAD28554.1; ScAcCCA and WP_011030299.1; ScAcCCE) to identify genes coding for AcCCase from a thermophilic bacterium. Thereby, we identified a 3 ORF-containing operon from T. fusca YX, which appears to encode for subunits of a biotin-containing enzyme. The potential catalytic functionalities of the T. fusca subunits were computationally identified by the sequence alignments. These alignments compared the T. fusca sequences to homologous proteins whose catalytic capabilities have previously been biochemically well characterized; for example enzymes from E. coli (Guchhait et al., 1974; Li and Cronan, 1992a; Cronan and Waldrop, 2002), Arabidopsis (Nikolau et al., 2003; Sasaki and Nagano, 2004; Salie and Thelen, 2016), human (Huang et al., 2010; Tong, 2013), S. coelicolor (Rodríguez and Gramajo, 1999; Diacovich et al., 2002; Gago et al., 2011) and M. tuberculosis (Cole et al., 1998; Li et al., 2005; Gago et al., 2006).

Based on these sequence similarities, the three T. fusca ORFs where characterized as: (a) Tfu_2557 (WP_011292980) encoding the TfAcCCA subunit (62.1 kDa) that appears to carry the BC and BCCP functional components; (b) Tfu_2555 (WP_011292978) encoding the TfAcCCB subunit (57.6 kDa) that carries the CT functional component; and (c) Tfu_2556 (WP_011292979) encoding the TfAcCCE subunit (8.7 kDa), a non-catalytic structural component. More specifically, the T. fusca AcCCB subunit appears to encompass the two domains that are homologous to the split CT-β and CT-α subunits of the E. coli (Li and Cronan, 1992b; Cronan and Waldrop, 2002) and plant heteromeric ACCase (Sasaki et al., 1993; Shorrosh et al., 1996; Ke et al., 2000); the CT-β and CT-α domains occur sequentially at the N-terminal and C-terminal halves of the TfAcCCB sequence. Similarly, for TfAcCCA, the BC and BCCP components can be identified as N-terminal and C-terminal domains of this protein. A significant amount of sequence similarity can also be found between T. fusca AcCCE ORF and the homologous proteins of ScAcCCE (~54% similarity) and ScPccE (~61% similarity) from S. coelicolor.

The order of these three ORFs in T. fusca genome is TfAcCCB-TfAcCCE-TfAcCCA, and this parallels the order of these ORFs in Nocardiopsis gilva (Supplementary Figure 1) (Li et al., 2013). The genus Nocardiopsis is affiliated with the phylum Actinobacteria, and these species display wide ecological versatility and metabolic capabilities to generate a diverse set of bioactive secondary metabolites. N. gilva is a halophilic bacterium, isolated from hypersaline soils, which has the ability to produce antifungal, antibacterial, and antioxidant metabolites (Tian et al., 2013). Although the broader genetic synteny appears similar among Actinobacteria, with the three AcCC ORFs adjoining the birA gene that encodes the enzyme that biotinylates the AcCCA subunit, the specific context appears to be more variable among this phylum, with the additional genes located between the AcCCE and AcCCA loci of the S. coelicolor and M. tuberculosis genomes (Supplementary Figure 1).

The sequences coding for the T. fusca AcCCA, AcCCB, and AcCCE ORFs were recombinantly expressed in E. coli either individually or in different combinations in order to reconstitute and optimize the reassembly of the holoenzyme complex. In these experiments only one of the subunits was His-tagged at the N-terminus, which facilitated affinity purification with a Ni-NTA column of any complexes that formed among the co-expressed subunits. Initial expression optimization experiments revealed that the co-expression of TfAcCCE with TfAcCCB generated higher titers of the latter catalytic subunit, suggesting the stabilization of TfAcCCB by potentially being assembled into a complex with TfAcCCE. In these co-expression experiments, we separately confirmed the expression of the TfAcCCE subunit by expressing an S-tagged version of this subunit and detecting its expression by western blot analysis using the antibody against the S-tag (Figure 1A).

Following these initial expression optimization experiments, we developed an expression system in which the TfAcCCase holoenzyme was reconstituted in vivo by co-expressing the non-tagged TfAcCCA and TfAcCCE subunits, and simultaneously co-expressing a His-tagged TfAcCCB subunit. Figure 1B shows that affinity chromatography with a Ni-NTA column of the His-tagged TfAcCCB subunit led to the co-purification of TfAcCCA and TfAcCCE subunits, indicating that a holoenzyme complex was reconstituted and purified. Similar analyses with His-tagged TfAcCCA and His-tagged TfAcCCB, indicate that TfACCE associates with the latter subunit, but not the former (Figure 1C).

Gel filtration chromatography was used to determine the molecular weight of the complexes that formed when different combinations of TfAcCCE, TfAcCCB and/or TfAcCCA were co-expressed and co-purified (i.e., those analyzed by SDS-PAGE as shown in Figure 1). Co-expressing all three subunits results in the recovery of complexes that resolved as two major peaks with apparent molecular weights of 1,160 ± 45 kDa (peak i) and 880 ± 30 kDa (peak ii), and a minor peak at 350 ± 13 kDa (peak iii) (Figure 2A); all three peaks contained the entire complement of the three subunits (Supplementary Figure 2). The molecular weight of these complexes was determined at three different protein concentrations, each differing by 2-fold. These molecular weight determinations differed from each other by <5% (data not shown), indicating that each of these peaks represent stable holoenzyme complexes of different subunit stoichiometry. As demonstrated in Figure 1B, SDS-PAGE analyses indicate that in these preparations the TfAcCCA and TfAcCCB subunits occur in an equal molar ratio. Additionally, densitometric analysis of these gels indicate that the staining intensity of the TfAcCCE subunit band occurs at 6.6% (±0.5) of the intensity of both the TfAcCCA and TfAcCCB subunits. This ratio is close to the molecular weight ratio of these subunits [8.7 kDa:(62.1 kDa + 57.6 kDa) = 0.073], indicating that the 3 subunits occur at equal molar ratio. [Note, we determined the ratio of E:(A + B), rather than E:A:B, because gels had to be overloaded in order to visualize the AcCCE-subunit, which caused the merger of the AcCCA and AcCCB subunits bands, making it difficult to separately conduct densitometry of the A and B subunit]. In combination therefore, we surmise that these 3 gel-filtration elution peaks (i, ii, and iii) may be dimeric and trimeric forms of an A₃B₃E₃ monomeric unit, with a preference for the formation of the dimer, (A₃B₃E₃)₂ that has a molecular weight of ~880-kD.

Analogous experiments were conducted to evaluate the role of TfAcCCE in the formation of potential subcomplexes with either TfAcCCA or TfAcCCB. The co-expression of His-tagged TfAcCCB with TfAcCCE resulted in the recovery of a subcomplex that included both subunits, whereas such a heteromeric subcomplex was not formed when His-tagged TfAcCCA was co-expressed with TfAcCCE. Based on the apparent molecular weight of the major peak recovered from the TfAcCCB-TfAcCCE subcomplex (349 ± 24 kDa; peak ix) (Figure 2C), and assuming that the two subunits occur in equimolar ratio, as occurs in the holoenzyme complex, the oligomeric state of this subcomplex would be B₆E₆. Less abundant peaks were also recovered in these preparations (peak vii at 1,199 ± 91 kDa and peak viii at 756.3 ± 44.7 kDa) (Figure 2C), which may therefore be higher oligomeric states, possibly (B₆E₆)₂ and (B₆E₆)₃, respectively. Although TfAcCCA does not appear to associate with TfAcCCE, it does form a series of subcomplexes with apparent molecular weights of 168 ± 10 kDa, 353 ± 18 kDa and 1,232 ± 69 kDa (peaks xii, xi, and x, Figure 2D); the preferred subcomplex (peak xii) maybe trimeric (A₃) and the two minor peaks being higher multimers of this trimeric sub-complex [i.e., (A₃)₂ and (A₃)₆, respectively].

The catalytic activity of the TfAcCCase holoenzyme was assayed and compared to that of the TfAcCCE-deficient enzyme (Figures 3A,B; Table 1). These experiments were conducted by first purifying TfAcCCB from a strain that was also co-expressing TfAcCCE, or from a strain that was expressing TfAcCCB in the absence of TfAcCCE. These two preparations were mixed in vitro with purified TfAcCCA and assayed for catalytic activity with the three potential acyl-CoA substrates, acetyl-CoA, propionyl-CoA or butyryl-CoA. These data indicate that although the reconstituted TfAcCCE-deficient enzyme is catalytically competent, the presence of TfAcCCE subunit increases the specificity constant (V_max/K_m) of the reconstituted holoenzyme by about 20-fold. This enhancement of catalytic activity occurs with all three substrates that were evaluated, with propionyl-CoA being the preferred substrate, followed by acetyl-CoA and then butyryl-CoA. In all cases, the TfAcCCE-enhancement of catalytic efficiency was primarily due to increases in V_max, which was increased by 20- to 100-fold depending on the substrate, with only minor changes in K_m values for these substrates. These kinetic parameters of the TfAcCCase enzyme are comparable to those previously determined for similar enzymes isolated from S. coelicolor and M. tuberculosis (Supplementary Table 2).

The selection of the organic substrate that is carboxylated by biotin-dependent enzymes is primarily a characteristic of the CT catalytic functionality, which is the 2nd half-reaction catalyzed by these enzymes. Prior characterization of the PCCase and AcCCase from S. coelicolor has indicated that a conserved residue (D422 of ScPccB; I420 of ScAcCCB) situated at the C-terminal end of these B subunits influences the substrate specificity of these enzymes (Diacovich et al., 2004). Comparing the amino acid sequences of TfAcCCB with ScPccB and ScAcCCB, identified that residue D427 of the former protein corresponds to residue D422 and I420 of the latter two proteins. Additional sequence comparisons establish that this residue is conserved among biotin-dependent carboxylases that can catalyze the carboxylation of propionyl-CoA, but it is less conserved in the broader context of other CT catalytic subunits that carboxylate other acyl-CoA substrates (Figure 4).

Additional confirmation of the equivalence of D427 of TfAcCCB to D422 of ScPccB was gained by generating a computational model of the tertiary structure of the TfAcCCB subunit. This model was generated with I-TASSER (Roy et al., 2010; Yang et al., 2014; Yang and Zhang, 2015), which finds structural templates in PDB, using multiple threading alignment approaches. The resulting computational model was superimposed on the experimentally determined structure of S. coelicolor PccB (PDB: 1XNY). The two structures were aligned using PyMol (https://pymol.org/2) and they align with a root means square deviation (RMSD) of 0.3 Å (Figure 5). This tertiary structure alignment reinforced the identification of D427 of the TfAcCCB subunit as equivalent to D422 of ScPccB. Specifically, D422 of ScPccB (and D427 in the TfAcCCB model) is situated deep in the substrate binding pocket, and the side chain of this residue faces the acyl-group of the acyl-CoA substrate.

We evaluated if mutating the conserved D427 residue of TfAcCCB to Ile will affect the substrate specificity of TfAcCCase, as occurs with ScPCCase and ScAcCCase. Initial characterization of the purified TfAcCCase, assembled with the TfAcCCB-D427I mutant subunit established that the major form of the recovered holoenzyme has a molecular weight of 897 ± 39 kDa, which is indistinguishable from that of the wild-type enzyme (Figure 2B). This mutation therefore, did not affect the ability of the enzyme to assemble the normal quaternary organization. Moreover, this reconstituted TfAcCCase mutant enzyme demonstrates a relatively subtle change on substrate specificity. The TfAcCCB-D427I mutation increased the specificity constant (V_max/K_m) for the acetyl-CoA carboxylation reaction (by 150%) and decreased, by 40–80%, the specificity constant of the butyryl-CoA or propionyl-CoA carboxylation reactions. These alterations are primarily due to changes in V_max, whereas the K_m values for each substrate were unaffected (Table 2, Figure 6).