Crystallization and X-ray diffraction studies of a complete bacterial fatty-acid synthase type I

Bacterial and fungal type I fatty-acid synthases (FAS I) are evolutionarily connected, as bacterial FAS I is considered to be the ancestor of fungal FAS I. In this work, the production, crystallization and X-ray diffraction data analysis of a bacterial FAS I are reported.


Introduction
Fatty acids are important as principal components of cellular membranes, as post-translational modifiers of proteins, as a storage form of energy and as signalling molecules. The synthesis of fatty acids is performed in a catalytic cycle that is largely conserved in nature. Despite the conservation of the chemistry of synthesis, the structures of fatty-acid synthases (FAS) differ significantly in eukaryotes and bacteria. FAS are classified into type I and type II systems. Type I FAS are protein complexes of up to 2.7 MDa in size. The catalytic sites are embedded in an elaborate architecture and substrates are shuttled as covalently bound molecules to acyl carrier protein (ACP) domains. The other 'conventional' FAS system that is present in most bacteria and mitochondria is referred to as type II, in which separate monofunctional proteins perform the specific steps required for fatty-acid synthesis (Schweizer & Hofmann, 2004;Gago et al., 2011).
FAS type I multi-enzyme complexes have been extensively analyzed in recent years by X-ray crystallographic and cryoelectron microscopic (cryo-EM) studies (Maier et al., 2010;Grininger, 2014). Whereas the mammalian FAS I forms X-shaped homodimeric complexes, the microbial (CMNbacterial and fungal) FAS I occurs in barrel-shaped multimeric complexes with D 3 symmetry. Microbial FAS I is found in essentially two stoichiometries: an 6 6 heterododecameric complex occurring in most fungal FAS I and a homohexameric complex present in some fungi and in the CMN-group bacteria. The structure of the 2.6 MDa yeast FAS I has been determined by X-ray crystallography Lomakin et al., 2007;Johansson et al., 2008Johansson et al., , 2009Leibundgut et al., 2007). The 1.9 MDa bacterial FAS I structure was characterized by cryo-EM, and a structural model was obtained by docking a fungal FAS I homology model (Boehringer et al., 2013). Cryo-EM studies of the conformational dynamics additionally contributed to the structural knowledge of FAS I systems (Brignole et al., 2009;Gipson et al., 2010;Ciccarelli et al., 2013), particularly by showing that bacterial and mammalian FAS I are conformationally dynamic, which critically determines their molecular mode (Grininger, 2014).
The bacterial FAS I system is only found in the CMN-group bacteria (Gago et al., 2011;Figs. 1a and 1b). The most important representative of the FAS I-carrying bacteria is the highly pathogenic organism Mycobacterium tuberculosis, the causative agent of tuberculosis (TB). Functional studies have characterized M. tuberculosis FAS I as producing C 16 and C 26 fatty acids de novo, which serve as substrates for FAS type IImediated and Pks13-mediated synthesis of the branched mycolic acids, which can be up to 90 C atoms in length (Kikuchi et al., 1992;Bhatt et al., 2007). The relevance of mycolic acids as an efficient barrier for M. tuberculosis is best illustrated by the use of three inhibitors of mycolic acid biosynthesis in antibiotic therapy for TB, pyrazinamide (PZA), isoniazid (INH) and ethambutol (Ma et al., 2010), with PZA identified as specifically targeting M. tuberculosis FAS I (Sayahi et al., 2011;Zimhony et al., 2000;Fig. 1a).

Figure 1
Overview of microbial de novo fatty-acid synthesis. (a) Synthetic machineries for fatty-acid production in bacteria and fungi. The grey background highlights the compartmentalized synthesis of acyl-CoA by the multifunctional FAS I systems. Key steps in the catalytic cycle of FAS I-mediated synthesis are as follows. The central ketoacyl synthase (KS) domain condenses ac(et)yl with malonyl to form -ketobutyryl. Further, the compound is reduced by a ketoacyl reductase (KR) to -hydroxyacyl, dehydrated by a dehydratase (DH) to produce enoyl, and again reduced by an enoyl reductase (ER). Substrate CoA esters are turned into ACP esters by transferases. ACPs (boxes in magenta) then shuttle substrates and intermediates to the active sites of the catalytic domains. Note that in FAS I, ACPs are part of the multienzyme. Also note that MPT (malonyl palmitoyl transferase) in fungal FAS I loads malonyl and unloads palmitoyl, and MAT (malonyl acyl transferase) in bacteria loads malonyl and unloads acyl chains of various chain lengths. The terminal thiol group of ACP is indicated to highlight that substrates and intermediates are provided as thioesters. Product specificities for corynebacterial, mycobacterial and fungal FAS I systems are given as reported in this study (C. efficiens FAS I) and previously: C. ammoniagenes FAS I (Stuible et al., 1997), M. tuberculosis and M. smegmatis FAS I (Kikuchi et al., 1992;Zimhony et al., 2004;Peterson & Bloch, 1977) and Saccharomyces cerevisiae FAS I (Sumper et al., 1969;Kresze et al., 1977). Separate proteins of the FAS II are labelled as occurring in M. tuberculosis fatty-acid synthesis (Gago et al., 2011;Bhatt et al., 2007;Sacco et al., 2007). Pharmaceutically relevant inhibitors of M. tuberculosis fatty-acid synthesis are included in the scheme: pyrazinamide (PZA; Sayahi et al., 2011), isoxyl (ISO;Gannoun-Zaki et al., 2013), thiolactomycin (TLM; Kremer et al., 2000), isoniazid (INH), ethionamide (ETH) and triclosan (TRC) (Sullivan et al., 2006;Banerjee et al., 1994). We recently reported the cryo-EM structure of M. tuberculosis FAS I (Ciccarelli et al., 2013). For crystallographic studies, we identified Corynebacterium efficiens FAS I, which is 52% identical in sequence to M. tuberculosis FAS I, as a promising target. Here, we present the expression, purification, thermal stability data, crystallization and preliminary X-ray analysis of FAS I from C. efficiens.

Macromolecule production
The cloning and purification of M. tuberculosis FAS I (UniProt accession code P95029) and C. ammoniagenes FAS I (UniProt accession code Q04846) have been described in detail elsewhere (Kikuchi et al., 1992); i.e. cloning with an infusion cloning reaction (Takara Bio, Japan) and purification with engineered N-terminal Strep-II Tags (Schmidt & Skerra, 2007).
The coding sequence for C. efficiens FAS I (UniProt accession code Q8FMV7) was amplified by PCR from C. efficiens genomic DNA (DSM 44549). The gene was inserted into an SalI/XhoI-digested pME164 plasmid with an in-fusion cloning reaction (Takara Bio, Japan) to yield pME300 [pME164 is a pET-22b(+) (Merck, Germany) derivative with a N-terminal fused Strep-II Tag-encoding sequence]. Plasmid pME300 was modified using site-directed mutagenesis (forward primer GCG GCG CGC CGC AAC CAG CTG C; reverse primer GGT TGC GGC GCG CCG CGA CAC CCT CCA CGA GTG TC) to generate plasmid pME301 with S1778A and S1779A mutated ACP. Mutations research communications Table 1 Macromolecule-production information.   were introduced to produce a highly homogeneous batch of non-post-translationally modified protein.
For the cloning of AcpS from C. efficiens in pET-coco-I, the coding sequence was prepared via PCR from genomic DNA (DSM 44549; forward primer AGA AGG AGA TAT AAG CAT GAT CTC GAT TGG AAC CGA TC; reverse primer TCG AGT GCG GCC TAG G CTA CCT GTT CTC GGT GGC CAC). The pET-coco-I plasmid was digested with SphI/ AvrII and assembly was carried out with an in-fusion reaction to yield pME150.
Proteins were purified essentially as described elsewhere (Ciccarelli et al., 2013). Escherichia coli plasmids pME300 and pME301 were transformed into BL21 Gold (DE3) cells (Agilent Technologies, USA). 35 ml LB medium precultures were inoculated with single ampicillin-resistant colonies and, after overnight growth, were used to further inoculate 2 l TB medium cultures. 2 l cultures were grown to an OD 600 of 0.8-1.0 at 37 C and 180 rev min À1 and then cooled to 20 C and induced with a final concentration of 0.5 mM IPTG. After 16 h, the cells were harvested by centrifugation, frozen in liquid nitrogen and stored at À80 C. For purification, cells were resuspended in buffer W (0.1 M sodium phosphate pH 7.2, 0.15 M NaCl, 1 mM EDTA); protease inhibitor (Roche, Switzerland) and DNaseI (Applichem, Germany) were added before breaking the cells with a French press. Lysates were centrifuged for 1 h at 4 C at 45 000g. Supernatants were transferred onto a 10 ml affinity Strep-Tactin gravity-flow column (IBA, Germany), washed five times with one column volume of buffer W and eluted with 6 Â 0.5 column volumes of buffer E (0.1 M sodium phosphate pH 7.2, 150 mM NaCl, 1 mM EDTA, 2.5 mM desthiobiotin). Proteins were polished by size-exclusion chromatography (SEC) using an Ä KTAexplorer (GE Healthcare, USA) on a Superose 6 XK 16/70 column (GE Healthcare, USA) with buffer C (0.1 M bis-tris propane pH 6.8, 0.2 M NaCl, 10% glycerol). Macromoleculeproduction information is summarized in Table 1.
For the preparation of enzymatically active protein, C. efficiens FAS I was co-expressed with C. efficiens AcpS protein. Expression was essentially performed as described above, except that both plasmids pME300 and pME150 were transformed and positive clones were selected by ampicillin/ chloramphenicol double resistance.
Preparations of C. efficiens FAS I from co-expression with AcpS were tested for FAS activity in a NADPH consumption assay (Lynen, 1969). For the assay, in a total volume of 100 ml, 25 mg FAS I in buffer AB (0.4 M potassium phosphate pH 7.3, 3 mM dithiothreitol) was incubated with 30 nM NADPH (Sigma-Aldrich, USA) and 50 nM acetyl-CoA (Sigma-Aldrich, USA). The absorption was recorded at 334 nm (Lambda 35, PerkinElmer, USA) for approximately 1 min before 60 nM malonyl-CoA (Sigma-Aldrich, USA) was added to the solution (Fig. 2c).
For determination of the product spectra, reaction solutions with the above-described composition were incubated overnight at room temperature. After 16 h, C 17 -CoA (Sigma-Aldrich, USA) was added as an internal standard and the reactions were stopped by adding precooled (À20 C) acetone. These solutions were mixed for 20 s and stored for 60 min at À20 C. After centrifugation for 5 min at 20 000g, the supernatants were evaporated in a SpeedVac at 4 C. The CoA esters were dissolved in 60 ml Milli-Q water and, after ultrasonication, were analysed by HPLC-UV-MS [Ultimate 3000 RSLC (Thermo Fisher Scientific, USA) coupled to a micrOTOF-Q II system (Bruker Daltonics, Germany)]. CoA esters were separated on an RP-18 column (100 Â 2.1 mm, particle size 1.7 mm; Waters, USA) in gradients of solvent A (water, 10 mM triethylamine/acetic acid buffer pH 9.0) and B (acetonitrile). Data were analyzed using the DataAnalysis 4.0 software package (Bruker Daltonics, Germany).

Crystallization
Despite extensive crystallization trials, both M. tuberculosis and C. ammoniagenes FAS I gave crystals that diffracted to only about 8 Å resolution (data not shown). Hence, we chose another sequence homologue, FAS I from C. efficiens, for its reported moderate thermotolerance and the expected higher conformational stability of its proteins (Nishio et al., 2003). Initial crystallization conditions for C. efficiens FAS I were identified by carrying out screening trials using various commercial sparse-matrix crystallization screening kits (The AmSO4 Suite, The Cations Suite, The JCSG+ Suite, The JCSG Core I-IV Suites, The PACT Suite, The PEGs Suite and The PEGs II Suite from Qiagen, Index HT from Hampton Research, and Morpheus HT-96 and The PGA Screen from Molecular Dimensions) using the sitting-drop vapourdiffusion method in 96-well Greiner plates at 4 and 22 C. Tiny crystals were observed in The PACT Suite at 4 C after 2 d. This condition was refined by the hanging-drop vapourdiffusion technique in 24-well plates using 1 ml protein solution in the droplet mixed with 1 ml reservoir solution. The best crystals were obtained at 0.1-0.3 M sodium malonate, 0.1 M bis-tris propane pH 7.5, 17-22% PEG 3350 at 4 C. Crystals grew after about one week with sizes ranging from 100 Â 100 Â 75 to 150 Â 150 Â 75 mm (Fig. 2f ). Crystallization  information is summarized in Table 2. C. efficiens FAS I crystals were soaked with Ta 6 Br 12 (Jena Bioscience), W 12 (Alfa Aesar), W 18 (a gift from Professor Robert Huber) or Nanogold (Nanoprobes) for 10 min, 1 h, 2 h or overnight. The concentrations of the soaking solutions for all clusters were varied in the low-micromolar range. For co-crystallization, the heavy atoms were added to the abovementioned crystallization conditions to a final concentration that was in the lowmicromolar range. In both cases, the crystals showed poor diffraction (7 Å or worse) when exposed to X-rays and rapid decay in diffraction intensities after a few exposures. Data sets could not be collected.

Data collection and processing
C. efficiens FAS I crystals were transferred to a cryoprotectant solution consisting of 20% ethylene glycol in the crystallization buffer, picked up in a nylon-fibre loop and plunged into liquid nitrogen. All crystals were exposed to single-wavelength X-radiation on beamline ID14-4 at the European Synchrotron Radiation Facility (ESRF) and maintained at 100 K while data were recorded on a CCD detector (ADSC Quantum Q315r). Individual diffraction images from C. efficiens FAS I crystals were initially indexed with iMosflm (Battye et al., 2011) to determine the crystal form. Data were processed with the xia2 suite of programs (Winter, 2010) running the XDS package (Kabsch, 2010) and AIMLESS (Evans, 2006). Crystals gave two different crystal forms in space groups R32 and C2 with differing unit-cell parameters (Table 3), designated crystal forms I and II. All diffraction data are publicly available at Zenodo (http://dx.doi.org/ 10.5281/zenodo.20031).

Results and discussion
We have recently established protocols for the recombinant overexpression of M. tuberculosis and C. ammoniagenes FAS I. C. efficiens, which is evolutionarily related to C. ammoniagenes, is reported to be moderately thermotolerant (Nishio et al., 2003), and we expected a higher conformational stability of its protein inventory. Two FAS I-coding genes (FAS A, NP_737523.1, and FAS B, NP_739002.1) were identified in C. efficiens (Stuible et al., 1997). FAS B (UniProt accession code Q8FMV7) was picked as a target for our studies owing to its higher sequence identity to M. tuberculosis FAS I (44 and 52% for FAS A and FAS B, respectively). About 5 mg of protein per litre of E. coli expression culture was obtained after chromatographic purification (Fig. 2a). The hexameric conformation was monitored by size-exclusion chromatography (Fig. 2b), and catalytic activity was demonstrated by spectroscopically monitoring NADPH consumption during the reductive steps in fatty-acid synthesis. The specific activity was determined as 270 AE 55 mU mg À1 , which is comparable to previously reported values ( Fig. 2c; Stuible et al., 1997). Under the assay conditions, C. efficiens FAS I produced C 16 -CoA (86%) and C 18 -CoA (14%). The protein stability of C. efficiens FAS I, as well as the homologous FAS I from C. ammonia-genes and M. tuberculosis, was analyzed by the fluorescencebased thermal shift (Thermofluor) assay and CD spectroscopy. Melting points of 45.4 C for M. tuberculosis FAS I, 44.6 C for C. ammoniagenes FAS I and 47.3 C for C. efficiens FAS I were determined, indicating low thermal stabilities of bacterial FAS I (Figs. 2d and 2e; Ericsson et al., 2006). Aiming for the highest conformational homogeneity, the structure of C. efficiens FAS I was characterized on a mutant construct bearing the modifications S1178A and S1779A. These mutations prevent posttranslational phosphopantetheinylation of the protein and should avoid the binding of intermediates or substrates to ACP that could induce different conformational states (Whicher et al., 2014). This was carried out despite the observation that C. efficiens FAS I is inactive when expressed alone without C. efficiens AcpS.
Despite several optimization trials (in terms of protein purification, crystallization and post-crystallization protocols; e.g. the use of dehydration), diffraction data from the crystals could be acquired to only 4.5 Å resolution. Data sets from two crystals gave different unit-cell parameters in space groups R32 and C2, referred to as crystal forms I and II, respectively ( Table 3). The packing of the FAS molecule in the asymmetric unit of the crystal was analyzed using MATTHEWS_COEF (Matthews, 1968) from the CCP4 suite (Winn et al., 2011). For crystal form I, calculation of the Matthews coefficient (V M ) gave values of 4, 2 and 1.3 Å 3 Da À1 , corresponding to solvent contents of 69, 38 and 7.1% for one, two and three molecules per asymmetric unit, respectively. This indicated the presence of one molecule in the asymmetric unit and the absence of noncrystallographic symmetry. In case of crystal form II, up to six chains (a whole hexamer barrel) can be accommodated in the asymmetric unit, indicating clear noncrystallographic research communications  (Karplus & Diederichs, 2012) of 0.5. The mean I/(I) falls below 2.0 at 6.5 Å resolution. ‡ Data were scaled to a CC 1/2 (Karplus & Diederichs, 2012) of 0.5. The mean I/(I) falls below 2.0 at 4.7 Å resolution. symmetry. For crystal form II, search templates comprising the hexameric barrel, a trimeric dome, a dimer or a monomer individually did not yield unique solutions in the molecularreplacement analysis.
In the case of crystal form I, the phase problem could be solved by molecular replacement (MR) using the structural model of M. smegmatis FAS I (Boehringer et al., 2013; PDB entry 3zen) as a model in MOLREP (Vagin & Teplyakov, 2010) with the help of the CCP4 suite. The MR solution was confirmed with Phaser (McCoy et al., 2007), which gave an LLG score of 424 and a TFZ score of 26. It should be noted that the overall sequence identity between C. efficiens FAS I and M. smegmatis FAS I is 53% (with 67% positives and 4% gaps; data from BLAST; Altschul et al., 1990). A pairwise alignment between the protein sequences of M. smegmatis FAS I and C. efficiens FAS I was used in phenix.sculptor (Bunkó czi & Read, 2011) from the PHENIX package (Adams et al., 2010) to mutate the MR solution (PDB) file to the C. efficiens FAS I sequence. An immediate rigid-body refinement with the different domains (sequence ranges: AT, 10-367;ER, 368-879;SBS, 880-1002;DH, 1003DH, -1289MAT, 1290MAT, -1674DM, 1953DM, -2065KR, 2073KR, -2357KS, 2358-3022) as independent rigid groups using PHENIX gave R and R free values of 0.32 and 0.40, respectively. Although the diffraction data set is of low resolution, a preliminary view of the electron-density map displayed the different domains of this megadalton molecule. In the next step, we attempted model improvement using the phenix.morph_model command (Terwilliger et al., 2012) invoking the prime-and-switch map (Terwilliger, 2004), but did not observe any significant improvement in either the refinement (R/R free ) values or the electron density.