1.8 Å resolution crystal structure of the carbapenem intrinsic resistance protein CarF

Abstract

The natural production of the β-lactam antibiotic carbapenem in bacteria involves a group of enzymes that form a synthetic pathway as well as proteins that protect the cell from self-intoxification by the products. Here, the crystal structure of CarF, one of the two proteins that confer resistance to synthesis of the antibiotic in the host organism, is reported. The CarF fold places it within a widely occurring structural family, indicating an ancient structural origin from which the resistance function has been derived.

1. Introduction

The emergence and increasing prevalence of enterobacterial resistance to carbapenem antibiotics is a growing cause for concern in the medical, public health and scientific communities (McKenna, 2013). As ‘antibiotics of last resort’, the carbapenems play a key role in the treatment of lifethreatening, drug-resistant infections. In particular, the cooccurrence of carbapenem resistance with resistance to other, more commonly used, antibiotics is a worrying trend.

Carbapenem antibiotics are naturally occurring bicyclic β-lactams. The first carbapenem discovered was thienamycin (from Streptomyces cattleya) followed by 1-carbapen-2-em-3-carboxylic acid (from Erwinia/Pectobacterium among others). Although these molecules show antibiotic activity, they are made in low abundance by the producer strains, are readily hydrolysed by renal dihydropeptidases and are too unstable for clinical use (Moellering et al., 1989; Fukasawa et al., 1992). Chemically synthesized derivatives, including imipenem, meropenem and ertapenem, are significantly more stable and are used successfully to combat pathogens expressing extended-spectrum β-lactamases. To date, over 80 carbapenem-derived compounds, most with improved antimicrobial properties and clinical stability, have been described in the literature. However, few of these are currently used clinically (Papp-Wallace et al., 2011). As resistance to the clinically employed carbapenem derivatives continues to evolve under strong selection, and as carbapenem-resistance genes in pathogens are disseminated internationally, understanding of carbapenem-resistance mechanisms is becoming increasingly important.

Production of the simplest carbapenem, 1-carbapen-2-em-3-carboxylic acid, is limited to a small group of enterobacteria, including Serratia sp. ATCC 39006 (S. 39006), Photorhabdus luminescens TT01, Erwinia herbicola and various Erwinia carotovora subsp. carotovora (Ecc) strains (now reclassified as Pectobacterium carotovorum) (Parker et al., 1982; McGowan et al., 1998; Derzelle et al., 2002; reviewed in Whitehead et al., 2001). However, cryptic car gene clusters have been found in a wider range of bacterial strains (Holden et al., 1998). In Ecc and S. 39006, carbapenem production is controlled by the car operon, which includes five well characterized biosynthetic genes (carA, carB, carC, carD and carE), two carbapenem intrinsic resistance (CIR) genes (carF and carG) and a gene of unknown function (carH). Carbapenem production is controlled by a complex hierarchical regulatory network that includes quorum sensing (Coulthurst et al., 2005).

The control of the car cluster in both Ecc and S. 39006 has been studied extensively. For more information on car operon regulation in Ecc and S. 39006, see Coulthurst et al. (2005), Wilf et al. (2011), Poulter et al. (2011),Clifton et al. (2003), Bodner et al. (2011), Phelan & Townsend (2013) and Hamed et al. (2013). However, the intrinsic resistance mechanism encoded by carF and carG has been less well studied.

Previous work in Ecc has shown that CarFEcc and CarGEcc are highly conserved within the group of carbapenem-producing strains. They do not show homology to any proteins known to be involved in antibiotic resistance. CarF and CarG are not able to hydrolyse other β-lactam antibiotics and do not confer cross-resistance to the clinically employed carbapenem imipenem (McGowan et al., 1995,1996,1997). Mutagenesis of the resistance genes has shown that CarF and CarG act in concert, each individually conferring partial carbapenem resistance and having complementary and additive effects (McGowan et al., 1997). CarF and CarG have no known sequence similarity and their mechanism(s) of action is, as yet, unknown. Recent studies have shown structural similarities between CarG_S.39006 and the periplasmic lysozyme inhibitor PliI-Ah (Tichy et al, 2014). The structural similarity between lysozyme and penicillin-binding proteins suggests that CarG_S.₃₉₀₀₆-mediated carbapenem resistance may function via interaction with the molecular targets of β-lactam antibiotics, the penicillin-binding proteins (PBPs; Tichy et al, 2014).

Formally, there are several possibilities for CarF and CarG to mediate carbapenem resistance. The first possibility would be some alteration of the molecular targets, the PBPs. Secondly, CarF and CarG could modify the cell wall to limit access of the antibiotic to its target. Formally, these carba-penem intrinsic resistance (CIR) proteins might also confer carbapenem resistance by acting as specific β-lactamases (although unlikely), by providing or activating efflux exporters, or via the deletion or modification of porin proteins to reduce antibiotic import (Fisher et al., 2005; Drawz & Bonomo, 2010).

Resistance to β-lactam antibiotics in general, and carba-penem antibiotics in particular, has been studied extensively in recent years because of the serious clinical implications of these emerging resistance mechanisms (Nordmann et al., 2011; Nordmann & Poirel, 2013). Various resistance mechanisms in clinically important pathogens have been described, including the production of ‘carbapenemases’ (Nordmann & Poirel, 2013; Yang et al., 1990; Queenan & Bush, 2007), efflux (Bornet et al., 2003) and the loss and/or modification of porins (Tängdèn et al., 2013; Chow & Shlaes, 1991; Armand-Lefévre et al., 2003; Correa et al., 2013). However, new carbapenem-resistance mechanisms are likely to emerge owing to strong positive selection through widespread use of this class of antibiotics (Nordmann & Poirel, 2013; Walsh et al., 2011).

This study is the first description of the crystal structure of the carbapenem intrinsic resistance protein CarF_S.₃₉₀₀₆. The driver for this study was the hypothesis that illuminating the structural characteristics of the CIR-family proteins may help to further our understanding of the existing mechanisms of intrinsic carbapenem resistance. Although the genes encoding the intrinsic resistance mechanism have not yet appeared in clinical drug-resistant isolates, advances in our basic knowledge about CIR-family proteins may help to predict future mechanisms of emerging clinical resistance to this increasingly endangered class of antibiotics.

2. Experimental procedures

2.1. Overexpression and purification

Recombinant CarF_S.39006 with a C-terminal hexahistidine tag was expressed in S. 39006 under the control of the IPTG-inducible T7 promoter in plasmid pQE80oriT. The plasmids and primers used for cloning are summarized in Table 1. The protein was purified and crystallized according to the protocol previously published for CarG_S.39006 (Tichy et al., 2014), but with some modifications. In brief, CarFS.39006 was purified from clarified whole-cell lysate with Ni-NTA agarose (Qiagen) according to the manufacturer’s guidelines [lysis buffer: 50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, 0.01% Triton X-100 (pH 7.5)]. Elution fractions were pooled and dialysed overnight against low-salt buffer (20 mM NaCl, 50 mM Tris-HCl pH 7.5, 1 mM DTT, 0.01% Triton X-100) prior to further purification using a HiTrap Q Sepharose anion-exchange column (GE Life Sciences). Further purification was performed by gel-filtration chromatography of pooled, concentrated elution fractions using a Superdex 75 column (GE Life Sciences) in 100 mM NaCl, 50 mM Tris, 1 mM DTT, 0.05% Triton X-100 (pH 7.5). Detergent was removed from the final concentrated sample by dialysis with BioBeads SM-2 Adsorbent Beads (Bio-Rad) according to the manufacturer’s instructions. The hexahistidine tag did not affect protein function, as demonstrated by the E. coli ESS carbapenem lethality bioassay described previously (McGowan et al., 1997; see Supporting Information).

Table 1. Macromolecule-production information.

Cleavage sites are shown in italics for oligonucleotides and the added hexahistidine tag is underlined.

Source organism	Serratia sp. ATCC 39006
DNA source	Serratia sp. ATCC 39006
Forward primer	GATGAATTCATTAAAGAGGAGAAATTAACTATGT–      ATTAAAGAGGAGAAATTAACTATGTTGAAAAA–      TCGTATT
Reverse primer	GTCAAGCTTAGTGATGGTGATGGTGATGTTTATC–      GTGATGGTGATGGTGATGTTTATCATCGATCG–      GCCT
Cloning vector	pQE80oriT
Expression vector	pQE80oriT
Expression host	Serratia sp. ATCC 39006
Complete amino-acid sequence of the construct produced	EILPIDEVNVDGGDFYVGLVFGKEDYAAHANTHL–     TPFSIMRTEVTYHQYQALQAWAETRGYQISGG–     CNGATFEDCWPSEKDGGRHPVTNVSWWDAVIF–     ANALSAQHNLQPYYVTADGQALKIPPEEGTDR–     GIRENPQASGYRLPTLAEWQVAARGGNKGLSD–     GTYGSRYAGKDQPASVANLPVSGTQTFSTLPV–     ASKQPNSLGLYDMSGNVSEWLNENYAVKGGKK–     MYYFCGGSYMDRVGSLASCDVHTPGFAMSDIG–     FRLVRPIDDKHHHHHH

2.2. Crystallization, X-ray data collection, structure determination and refinement

The initial crystallization condition for CarF was identified using a commercial screen (Morpheus, Molecular Dimensions) by the sitting-drop vapour-diffusion method at 18°C. The final optimized crystals were obtained in a precipitant solution consisting of 0.1 M Morpheus Carboxylic Acids, 0.1 M HEPES/MOPS pH 7.5, 17.5% MPD, 17.5% PEG 1000, 17.5% PEG 3350, again using the sitting-drop vapourdiffusion method (Table 2). The crystals were flash-cooled in liquid nitrogen using crystallization mother liquor supplemented with 25% glycerol as a cryoprotectant. X-ray diffraction data were collected from a single crystal of CarF on beamline I04 at Diamond Light Source (DLS) to 1.8 Å resolution and were processed using iMosflm (Battye et al., 2011) and AIMLESS (Evans & Murshudov, 2013) in the CCP4 program suite (Winn et al., 2011). Initial phase estimates and electron-density maps were calculated by molecular replacement with Phaser (McCoy et al, 2007) using an ensemble of CarF models generated by the Robetta de novo proteinmodelling webserver (Raman et al., 2009) as initial search models (Kim et al., 2004). Structural refinement of the initial model was performed using restrained refinement in REFMAC5 with jelly-body refinement and anisotropic temperature factors (Murshudov et al., 2011), in combination with iterative model building with Coot (Emsley & Cowtan, 2004). Noncrystallographic symmetry restraints were not used during structure refinement. The quality of the final model was assessed with MolProbity (Chen et al., 2010) and the geometry and all-atom contacts were ranked in the 92nd percentile in comparison to structures of similar resolution (data-collection and refinement statistics are summarized in Tables 3 and 4).

Table 2. Crystallization.

Method	Vapour diffusion, sitting drop
Plate type	MRC 2-drop, 96-well
Temperature (K)	291
Protein concentration (mg ml-1)	10
Buffer composition of protein solution	100 mM NaCl, 50 mM Tris, 1 mM DTT pH 7.5
Composition of reservoir solution	0.1 M Morpheus Carboxylic Acids, 0.1 M HEPES/MOPS pH 7.5, 17.5% MPD, 17.5% PEG 1000, 17.5% PEG 3350
Volume and ratio of drop	0.4 μl, 1:1
Volume of reservoir (μl)	70

Table 3. Data collection and processing.

Values in parentheses are for the outer shell.

Diffraction source	Beamline I04, DLS
Wavelength (Å)	0.978
Temperature (K)	180
Detector	PILATUS pixel
Crystal-to-detector distance (mm)	190.5
Rotation range per image (°)	0.2
Total rotation range (°)	90.0
Exposure time per image (s)	0.2
Space group	P212121
a, b, c (Å)	49.47, 55.45, 176.20
α, β, γ (°)	90, 90, 90
Mosaicity (°)	0.16
Resolution range (Å)	36.13–1.80
Total No. of reflections	144044 (9084)
No. of unique reflections	44709 (2788)
Completeness (%)	97.5 (99.1)
Multiplicity	3.2 (3.3)
〈I/σ(I)〉	8.0 (2.0)
Rr.i.m.	0.12 (0.56)
Overall B factor from Wilson plot (Å2)	16.077

Table 4. Structure solution and refinement.

Values in parentheses are for the outer shell.

Resolution range (Å)	36.13–1.80
Completeness (%)	97.2
σ Cutoff	2.44
No. of reflections, working set	44709
No. of reflections, test set	2185
Final R cryst	0.137
Final R free	0.208
Cruickshank DPI	0.267
No. of non-H atoms
Protein	3983
Ion	2
Water	305
Total	4290
R.m.s. deviations
Bonds (Å)	0.019
Angles (°)	1.87
Average B factors (Å2)
Protein	24.99
Ion	21.23
Water	37.95
Ramachandran plot
Most favoured (%)	96.15
Allowed (%)	3.46

2.3. Carbapenem-resistance bioassay

Carbapenem resistance in the bacterial strains of interest was assayed using the E. coli ESS killing bioassay described previously (McGowan et al., 1997). In essence, the E. coli strain ESS is supersensitive to β-lactam antibiotics and therefore enables the detection of low concentrations of the carbapenem antibiotic made by the producer strain. In contrast, recombinant E. coli ESS expressing cloned carF and carG genes (for intrinsic resistance to the carbapenem) exhibit significantly decreased sensitivity to the antibiotic, as shown previously (McGowan et al., 1997).

2.4. Analytical gel filtration

CarF_S.39006 complex formation was assayed by analytical gel filtration as described by Whitaker (1963), using hexahistidine-tagged CarF_S.39006 purified as described above. Gel filtration was performed on an S-200 analytical gel-filtration column (GE Healthcare) and the elution volumes were compared with those obtained using standards of known molecular mass (Gel Filtration Markers Kit for Protein Molecular Weights 12 000–200 000 Da; Sigma–Aldrich).

3. Results

The crystallographic asymmetric unit contains two protomers of CarF related by a twofold noncrystallographic symmetry axis. Analysis of the oligomeric state using PISA (Krissinel & Henrick, 2007) indicates that the buried surface area formed by the dimer seen in the asymmetric unit is 2090 Ų (9.9% of the total surface area of the dimer). Although this interface is sufficiently extensive for the dimer to be stable in solution (Krissinel & Henrick, 2007), we propose that it is a non-biological artefact of the crystal lattice. Analytical gel filtration of CarF_S.39006 shows that the protein elutes at a volume corresponding to a molecular mass of 25 kDa, which is consistent with the predicted monomer of hexahistidine-tagged CarF, which has a molecular mass of 30.0 kDa (see Supporting Information).

The structure of CarF_S.39006 and its secondary-structural elements are depicted in Fig. 1. In general, CarF_S.₃₉₀₀₆ has low secondary-structure content (20% α-helix and 19% β-strand). The N-terminus is composed of several short β-strands (β1–β4). These form two separate antiparallel sheets which wrap around the surface of the C-terminal helices α3 and α4. On the opposite surface of the CarF_S.₃₉₀₀₆ monomer, helices α1 and α2 pack such that they cross at nearly right angles. The only other secondary-structural elements are four short β-strands which form a short four-stranded antiparallel sheet in the centre of the molecule (strands β10–β13). Sequence conservation of CarF and known functional homologues are shown in Fig. 1(c). Mapping the conserved residues to the surface of CarF reveals that amongst the general distribution of the position of the conserved residues, a clustered belt of conserved residues is present on one face of CarF, encompassing sheets formed of strands β1–β2 and β10–β12 and a large region of low secondary structure (Supplementary Fig. S3).

Figure 1. The structure of CarF_S.39006.

(a) Two orthogonal views of CarF depicted as a cartoon representation. The CarF monomer is coloured as a rainbow (blue to red from the N- to C-terminus) and secondary-structural elements are labelled. (b) Topology diagram of CarF_S.₃₉₀₀₆, with structural features coloured to correspond to (a). (c) Structure-based sequence alignment of CarF and its known functional homologues. CarF secondary-structure elements are indicated below the protein sequences. Amino acids are shaded in accordance with CarF sequence conservation. A black background indicates a fully conserved residue. The observed secondary-structural elements in the crystal structure of CarF are numbered sequentially. α-Helices are shown as blue cylinders, 3₁₀-helices as green cylinders and β-strands as blue arrows. Atomic coordinates and structure-factor files have been deposited with the PDB (PDB entry 5aoh).

The electron-density maps revealed the presence of a potential bound metal ion. The putative metal ion has octahedral coordination and involves the side chains of Glu43 and Glu213, carbonyl O atoms from Asp206, Gly209 and Val211, and a nearby water molecule (see Fig. 2b ). Given the octahedral coordination of the metal ion this was likely to be either sodium, potassium or calcium. When modelling this potential ion as water and running the ‘highly coordinated waters’ option in Coot, this atom was suggested to be either potassium or calcium. Elemental ion placement in phenix.refine suggested this ion to be calcium. However, given that both sodium and potassium salts were present in the crystallization solution as part of the Morpheus Carboxylic Acids mix, potassium was modelled into the observed electron density following refinement. The putative potassium is buried within the CarF_S.39006 monomer and may function to stabilize the loops within this region of low secondary structure. Both CarF_S.39006 protomers contain two intramolecular disulfide bonds (Cys67–Cys75 and Cys231–Cys245). Both of these disulfide links appear to stabilize sheet structures, with the paired residues 67 and 75 bridging strands 5 and 6, and residues 231 and 245 bridging strands 11 and 12 (see Fig. 2a ).

Figure 2. Structural features of CarF.

(a) Disulfide bonds. A monomer of CarF is shown as a blue cartoon. Residues involved in intramolecular disulfide bonds are shown as yellow sticks and are labelled. The putative potassium ion is shown as a red sphere. (b) Putative metal ion. Two orthogonal views of a putative potassium ion bound in the CarF structure. The potassium ion is shown as a purple sphere, a coordinated water molecule is shown as a red sphere and amino-acid residues involved in metal-ion coordination are shown as sticks. The 2F_o — F_c electron-density map for this region is shown as a grey mesh contoured at 1σ, and the F_o — F_c map calculated with the potassium ion omitted is shown as a green mesh contoured at 3σ.

Structural homologues of CarF_S.39006 were identified via the DALI protein structural database server (Holm & Rosen-ström, 2010). The top-scoring structural matches for CarF_S.39006 consisted primarily of proteins with an FGE domain (formyl-glycine generating enzymes). CarF_S.39006 superimposes onto the human FGE protein with a core r.m.s.d. of 1.43 Å over 176 aligned residues (30% amino-acid sequence identity; PDB entry 1y1j; Dierks et al., 2005) and confirms that CarF does indeed consist of an FGE-like domain beginning at Pro36 and continuing to the C-terminus of CarF_S.39006 (Fig. 3a ). Using the amino-acid sequence of CarFS.39006 as a query for an iterative hidden Markov motif search with the Jackhmmer web server (Murzin et al., 1995), the Pfam domain FGE-sulfatase (PF03781.14) was identified spanning residues 27–286 of the input sequence. Further analysis with Jackhmmer reveals 2439 sequences in the Pfam database containing the FGE-sulfatase domain, often as part of a multidomain protein. The FGE domain is a subtype of the C-type lectin (CLec) fold (Finn et al., 2015) and it has been proposed that like the Ig fold and the leucine-rich repeat fold, the CLec fold can be considered as a stable structural template to allow massive sequence variation in enzymes of varying function (Coq & Ghosh, 2011). The ability of such proteins to accommodate such massive sequence variability has been proposed as a mechanism for these proteins to bind to novel ligands, and may in part explain the diversity of proteins that this domain is found in. In the CarF_S.₃₉₀₀₆ structure the N- and C-termini of the CLec domain form β-strands that pair with one another (strands β4 and β14). ^In CarF_S.39006 there is an N-terminal extension to the C-Lec domain. This extension wraps around the surface of the CLec domain and may serve to mask hydrophobic residues in this region. Residues 108–141 of CarF_S.39006 correspond to a region which is variable within FGE proteins. In TvpA of Treponema denticola this variable region appears to stabilize helix α1, and this region may play a similar role in CarF_S.₃₉₀₀₆ (Coq & Ghosh, 2011).

Figure 3. The FGE active site is not conserved in CarF_S.39006.

(a) Two orthogonal views of the structurally conserved regions of CarF (green) overlaid on human FGE (blue). (b) Two orthogonal views of the entire CarF structure (green) overlaid on human FGE (blue, with the active-site cysteines as yellow spheres). The orientation of the views are as in (a) and the boundary of the structurally conserved core is shown as a semitransparent surface. (c) Electrostatic surfaces of hFGE (left) and CarF (right) in the same orientation. The FGE active-site channel highlighted by the yellow dashed box (with the active-site cysteine in yellow) is not present in CarF.

FGE-type enzymes act by oxidizing cysteine or serine to produce formylglycine, which is a critical active-site residue of sulfatases. The active site of FGE proteins bears a groove containing proline and reactive cysteine residues, which are critical for catalysis (Dierks et al., 2005). A comparison of CarF_S.39006 with the catalytically active human FGE protein is shown in Fig. 3(b), with the active-site cysteine highlighted. The overlay shows that although the general FGE-domain structure is conserved in CarF_S.39006, the region around the active site is not. The substrate-binding groove present in human FGE appears as a flat surface in the CarF_S.39006 structure and, more importantly, the reactive cysteine is absent. Taken together, these data would suggest that CarF does not function as a bona fide formyl-glycine generating enzyme, despite the structural similarities.

4. Discussion

In this study, we describe the X-ray crystallographic structure of CarF_S.39006, a member of the CIR family of proteins. Earlier structural analyses of the co-operonic CarG_S.39006 showed structural similarities to prokaryotic lysozyme inhibitors (Tichy et al., 2014). Here, we find that CarF_S.39006 also shows structural homology to known proteins but has not retained the relevant catalytic residues. The precise mechanism of action of CarF_S.39006 in carbapenem resistance has yet to be elucidated, but is likely to include regulatory interaction with a partner protein or other macromolecule.

In light of the known homologies, it has been postulated that CarG_S.39006 may act by interaction with the molecular targets of carbapenem antibiotics, namely the penicillin-binding proteins (Tichy et al., 2014). Recent, preliminary research conducted in this laboratory has suggested that CarF_S.39006 may interact with outer membrane porins in Gramnegative bacteria. Consequently, it is tempting to speculate that the additive resistance-conferring functions of CarF_S.39006 and CarG_S.39006 in conferring resistance to carbapenem may be owing to distinct, yet complementary, mechanisms. This could be mediated by the protection, modification or sequestration of the target molecule by CarG coupled with the inhibition or diminution of antibiotic uptake through porin modification by CarF. While it has not yet been possible to determine the exact molecular mechanism by which CarF and CarG confer resistance to the carbapenem, the structural data identified a number of interesting sites for further mutagenesis studies.

In the face of the continuing emergence of clinically relevant carbapenem-resistant strains, understanding how naturally occurring carbapenem-resistance mechanisms have evolved may be increasingly relevant, particularly as clinically relevant drug resistance is known to be at least partially driven by horizontal gene transfer from nonpathogenic strains (Nordmann & Poirel, 2013; Walsh et al., 2011). The structural determination of the two essential carbapenem-resistance proteins will contribute to a broader understanding of enterobacterial carbapenem resistance.

Supplementary Material

Supporting information: this article has supporting information at journals.iucr.org/d