As part of a programme of synthesizing and investigating the biological properties of new fluoroquinolone antibacterials and their targeting of topoisomerase IV from Streptococcus pneumoniae, we have solved the X-ray structure of the complexes of two new 7,8-bridged fluoroquinolones (with restricted C7 group rotation favouring tight binding) in complex with the topoisomerase IV from S. pneumoniae and an 18-base-pair DNA binding site—the E-site—found by our DNA mapping studies to bind drug strongly in the presence of topoisomerase IV (Leo et al. 2005 J. Biol. Chem. 280, 14 252–14 263, doi:10.1074/jbc.M500156200). Although the degree of antibiotic resistance towards fluoroquinolones is much lower than that of β-lactams and a range of ribosome-bound antibiotics, there is a pressing need to increase the diversity of members of this successful clinically used class of drugs. The quinolone moiety of the new 7,8-bridged agents ACHN-245 and ACHN-454 binds similarly to that of clinafloxocin, levofloxacin, moxifloxacin and trovofloxacin but the cyclic scaffold offers the possibility of chemical modification to produce interactions with other topoisomerase residues at the active site.
Quinolone agents, which target type II topoisomerases in Gram-negative and Gram-positive bacteria, are very important drugs in our armoury for the treatment of serious microbial infections and for which there is increasing resistance to β-lactam antibiotics. Quinolone drugs are not exempt from resistance, but the levels of resistance exhibited by this class are at present lower than those for β-lactam-based antibiotics and macrolides (statistics from Canadian Bacterial Surveillance, 2009 ). Hence, there is a need to develop effective rationally designed new pharmaceuticals of this class in order to bypass resistance mutations. We have developed 7,8-bridged fluoroquinolones with the aim that the restricted movement of the C7 group should favour tight binding. Quinolones act on type II topoisomerase–DNA complexes to prevent strand passage of the T-segment through the cleaved G-gate in the topoisomerase cycle . When not interrupted by drug interaction, this cycle starts with the reversible binding (through the catalytic tyrosines) of the G-segment DNA to the G-gate consisting of the TOPRIM and WHD domains of the core complex of the type II topoisomerase (figure 1). Capture of a transported T-segment DNA duplex by closure of the N-gate ATPase domains allows strand passage through the double-stranded break in the G-gate. Religation of the cleaved G-DNA segment and exit of the T-DNA duplex through the C-gate completes the passage of one DNA through the other, and after ATP hydrolysis, the enzyme resets for another cycle (figure 1). Quinolones trap the covalent enzyme–DNA cleavage complex, preventing DNA resealing and triggering cell death.
In prokaryotes, DNA topology is controlled by two type II topoisomerase paralogues, topo IV (a tetramer with subunits ParE2ParC2) and gyrase (a tetramer with subunits GyrA2GyrB2). The former mediates decatenation of replicated DNA molecules and the latter regulates the overall level of DNA supercoiling within the cell, which is perturbed by polymerases, helicases and other enzymes involved in replication, transcription and recombination [2–14]. The elucidation of the structures of topoisomerase II–drug–DNA complexes at a resolution of 3.5 Å and above by ourselves [15–17] and other workers in the field has meant that there is now a large number of structures deposited with the PDB and a body of knowledge of how a wide range of quinolone and other drugs semi-intercalate into the DNA. How members of the spiropyrimidinetrione compounds [18–20] and also how quinazolinedione compounds [16,21] make contact with the DNA bases and surrounding amino acids residues of their topoisomerase IIA binding sites has been revealed [21–26]. Recent findings of Chan et al. show how the spiropyrimidinetrione analogues and etoposide bind differently in the hemi-intercalated binding site in the gyrase complex . We [16,27] and more recently other groups [21,28] have investigated quinazolinediones, a class of antimicrobials that bind through different side-chain interactions to quinolones and select different gyrase and topoisomerase IV mutations in the potential binding volume. Apparently minimal changes in drug structure lead to a lack of cross-resistance with quinolones, an encouraging outcome that can be rationalized structurally and potentially optimized for drug development.
A number of Gram-positive and Gram-negative pathogens have been identified by the Centers for Disease Control, USA as serious antimicrobial resistance threats, of which Streptococcus pneumoniae is a major medical concern. Streptococcus pneumoniae is a Gram-positive bacterium and an important human pathogen that causes a range of infections, including pulmonary pneumonia, meningitis and otitis. Certain strains have developed resistance to beta-lactams and erythromycin. As part of our ongoing structural and biochemical studies aimed at new topoisomerase-targeting therapeutics, we have focused on topoisomerase IV–DNA complexes with ACHN-245 and ACHN-454, two novel 7,8-bridged fluoroquinolones developed and produced by Achaogen (figure 2c). For comparison, figure 2 also shows structures of levofloxacin and moxifloxacin, two clinically important anti-pneumococcal fluoroquinolones, and of the potent investigational quinolones clinafloxacin and trovafloxacin. The new compounds are similar to clinafloxacin in having cyclopropyl and 3-aminopyrroin-1-yl groups at the respective 1- and 7-positions of the quinolone ring (figure 2). However, they differ in having a heptacyclic ring system formed between the quinolone 8-position and the 2-position of the 3-aminopyrrolidin-1-yl substituent at position 7; this seven-membered ring contains a double bond in ACHN-245 and a cyclic ether in ACHN-454. Levofloxacin has, by contrast, a hexacyclic ring system between positions 1 and 8 of the quinolone (figure 2). The attraction of incorporating these bridging ring systems is that they constrain rotation and provide a scaffold for drug engineering, two features that potentially increase binding affinity and selectivity. Here, we describe the enzyme inhibitory and microbiological activities of the compounds. Moreover, we report the structures of the topoisomerase IV–DNA cleavage complexes with ACHN-245 and ACHN-454 at 3.43 and 3.24 Å resolution, respectively.
As in previous structures, we have made use of a DNA duplex corresponding to the E-site, which we originally identified as an S. pneumoniae chromosomal DNA sequence that is strongly cleaved by S. pneunomiae topo IV (and gyrase) in the presence of a variety of different fluoroquinolones [29,30]. This property makes it an ideal DNA partner for crystallographic work on new quinolones. In contrast with the fluoroquinolone complexes previously published by our group, the E-site DNA oligomers used in this study were changed from 34mers to 18mers (figure 2b), as part of an optimization of experimental conditions which has yielded reliable crystallization and structure solution of DNA cleavage complexes. The shorter DNA substrate also produced a change in the crystal space group from the previously reported P32 to P3121 and has allowed higher-resolution structures of the complexes with quinolones to be determined via co-crystallization compared with those obtained by soaking. In turn, it has yielded better-resolved electron density for the chelated magnesium ions, which mediate key interactions among protein, DNA and quinolones in the topoisomerase IV cleavage complex [17,31].
2. Material and methods
2.1. Cloning, expression and purification of Streptococcus pneumoniae ParC and ParE proteins
The cloning, expression and purification protocols for S. pneumoniae topoisomerase IV ParC, ParE, ParC55 and ParE30 proteins have been described previously [16,32]. Quinolone-resistant ParC S79F and quinazolinedione-resistant ParE E475A proteins were overexpressed and purified as described [27,33,34].
2.2. Preparation of the DNA substrate
In order to form the S. pneumoniae cleavage complex, E-site 18mer DNA oligomers (5'-CATGAATGACTATGCACG-3', 5'-CGTGCATAGTCATTCATG-3') were synthesized by solid-phase phosphoramidite chemistry and doubly HPLC purified by Metabion, Munich. The lyophilized DNA oligomers were reconstituted in the annealing buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 5 mM β-mercaptoethanol, 0.05% NaN3), mixed in 1 : 1 molar ratio, heated to 95°C and then allowed to cool slowly to 4°C over a period of 48 h in a Dewar (sealed thermos flask). Macromolecule information is provided in table 1.
Topoisomerase IV core was formed by mixing ParC55 and ParE30 at an equimolar ratio in higher salt buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1 mM β-mercaptoethanol, 0.05% NaN3). The protein was then dialysed into lower salt buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM β-mercaptoethanol, 0.05% NaN3) and the cleavage complexes were formed by mixing topo IV and E-site 18mer DNA at 1 : 1.2 molar ratio, respectively. Magnesium chloride and the drug of interest were added to final concentrations of 10 mM and 1 mM, respectively. The complexes were allowed to form by incubation at room temperature overnight.
Crystallization information is summarized in table 2. Crystals of the cleavage complexes were obtained by vapour diffusion using the sitting drop technique in MRC Wilden crystallization plates. Drops were formed at 600 : 400 nl ratio for complex and precipitant solution, respectively, using a Mosquito robot from TTP Labtech (www.ttplabtech.com). The crystallization was performed using a gradient grid varying pH from 6.0 to 7.0, NaCl from 100 to 140 mM and isopropanol from 4 to 7%. The rest of the crystallization cocktail was kept constant (i.e. 50 mM Na cacodylate, 2% Tacsimate; Hampton Research) . Crystals of varying quality appeared stochastically throughout the gradients used and no clear indication of one strictly preferred crystallization condition was found within the boundaries of the grid employed. Significant prep-to-prep variation was also observed when different batches of the protein were used. Hence, it was found to be essential to scan a range of conditions each time the crystallization was performed. The best crystals were selected, briefly placed into a cryoprotectant solution (50 mM Na cacodylate, pH 6.5, 2% Tacsimate , 62.5 mM KCl, 7.5 mM MgCl2, 1 mM β-mercaptoethanol, 30% (v/v) MPD) and then frozen directly in liquid nitrogen.
2.3. Crystallization and data collection
Data collection and processing are summarized in table 3. Data collection was performed using the GDE software available on the I02 beamline at Diamond SLS, Oxfordshire. The programs XDS and XSCALE [36,37] were used for data reduction. Structure solution and refinement are summarized in table 4. The structures were solved with PHASER  as implemented in the CCP4 suite . Our 3K9F structure was used as the starting protein model. Refinement was performed with PHENIX [40,41] using multiple rounds of coordinate and temperature factor refinement (employing TLS and secondary structure geometry restraints). Manual model fitting and correction was performed in WinCoot [42,43].
Model quality was assessed using the in-built analysis tools of WinCoot (including geometry analysis and Ramachandran plot) as well as the verification tools provided by the PDB. Figures for this paper were generated using PyMOL , ChemDraw  and Corel Draw (www.coreldraw.com). The structure was verified using WinCoot and ProCheck .
2.4. Drug susceptibilities and topoisomerase assays
Bacterial susceptibility to drugs was determined by the broth microdilution assay following CLSI-recommended guidelines. Briefly, approximately 104 CFU of the S. pneumoniae strains were inoculated in a final volume of 100 µl cation-adjusted Mueller–Hinton broth supplemented with lysed horse blood and twofold dilutions of the test compounds in 96-well plates. Plates were incubated at 35°C for 20–24 h. The MIC is the drug concentration at which no growth was seen when tested under these conditions.
Methods for assaying DNA gyrase and topoisomerase IV activity including DNA cleavage have been described previously . Briefly, DNA cleavage assays were set-up using supercoiled pBR322 DNA (400 ng) as substrate. Full-length topo IV reconstituted by combining ParC (450 ng) and ParE (1 µg), or topo IV core domain ParE30-ParC55 fusion protein (400 ng), were incubated with DNA. Reaction buffer contained 40 mM Tris-HCl, pH 7.5, 6 mM MgCl2, 10 mM DTT, 200 mM potassium glutamate and 50 µg ml−1 BSA in a final volume of 20 µl with or without drug included. Samples were incubated at 37°C for 1 h followed by addition of 2 µl of 10% SDS to each reaction to induce DNA cleavage. After addition of proteinase K to 200 µg ml−1, incubation was continued at 42°C for 1 h to digest DNA-bound protein. Sample loading dye (5 µl) was added to each tube and DNA cleavage products were separated by gel electrophoresis in 1% agarose. DNA bands were stained with ethidium bromide and photographed under UV illumination. CC25 is the drug concentration used in the DNA cleavage assay that converted 25% of the supercoiled DNA substrate into linear DNA.
3. Results and discussion
We have co-crystallized S. pneumoniae topoisomerase IV ParC–ParE breakage-reunion domain (ParC55, residues 1–490) and ParE TOPRIM domain (ParE30, residues 390–631) with an 18 bp DNA duplex (the E-site) stabilized by the novel drugs ACHN-245 and ACHN-454 synthesized by Achaogen . The X-ray crystal structures of the complexes with ACHN-245 and ACHN-454 were determined at 3.43 and 3.24 Å, respectively, showing a closed ParC55 dimer flanked by two ParE30 monomers (figures 3–6). The macromolecular structure of this tetrameric complex is similar to that found for other S. pneumoniae topoisomerase–DNA–drug core complexes that we have previously reported [15–17]. We note that residues 6–30 of the N-terminal α-helix, α1 of the ParC subunit embrace the ParE subunit, pulling the ParE subunits close to either side of the ParC dimer (figure 3) . This interaction, absent from our original ParC55 dimer structure [48,49], appears to be very important for ParC–ParE complex stability as deletion of the α-1 arm resulted in loss of DNA cleavage activity . Figure 2a outlines the modular structure of the ParC and ParE proteins, mapping the positions of the TOPRIM metal binding domain, WHD (the winged helix domain)  and the TOWER regions of the protein subunits. This architecture has now been found in both topoisomerase IVs and gyrases from bacteria, and in topo IIα and topo IIβ from eukaryotes. The upper part of the topoisomerase complex consists of the E-subunit TOPRIM domain formed of four parallel β-sheets and surrounding α-helices. The WHD within the C-subunit together with the TOWER forms the U-groove-shaped protein region into which the G-gate DNA binds inducing a banana-shaped bend. The lower C-gate region (figure 3a) forms the region of the structure through which the T-segment DNA is envisaged to exit (see figure 1 illustrating the stages in the catalytic cycle) and is formed of a pair of two long α-helices terminated by a spanning short α-helix. Dimerization of the C domains at this point forms the 30 Å cavity, which will accommodate a B-DNA helix. The topoisomerase IV from S. pneumoniae is thought to follow the generic topoisomerase catalytic cycle shown in figure 1 involving putative intermediates 1–4, for which we have confirmation of intermediate 1 from our recent structure of the full complex (the holoenzyme less the CTD β-pinwheel domain) with the ATPase domains in the open conformation . Moreover, a closed state with dimerized ATPase domains has been observed in a yeast topo II structure from Berger's group .
The G-gate DNA for our S. pneumoniae complexes stabilized by 7,8-bridged quinolones is the 18 bp E-site sequence, our notation for a DNA site that we first found in cleavage mapping studies of the S. pneumoniae chromosome [29,30]. It is clear from the presence of covalent DNA–protein links to ParC Y118 residues that the protein structures solved here (and earlier [15,16]) represent quinolone-stabilized cleavage complexes formed by turning over the topoisomerase IV tetramer bound to DNA. Interestingly, the E-site DNA sequence has subsequently been used in the successful co-crystallization of quinolone–DNA cleavage complexes of Acinetobacter baumannii topoisomerase IV  and Mycobacterium tuberculosis DNA gyrase , though without identification or reference to its origins in the S. pneumoniae system. It appears that the E-site is a versatile DNA substrate that permits the crystallization and structure solution of complexes formed with a variety of type II topoisomerases and cleavage-enhancing drugs.
Within the S. pneumoniae topoisomerase IV complex, a 7,8-bridged quinolone is hemi-intercalated into each DNA strand and stacked against the DNA bases at the cleavage site (positions −1 and +1 of the 4 bp staggered cut in the 18mer DNA) (figures 3–6) as found now in many other Gram-negative and Gram-positive topoisomerase IV and gyrase complexes. Figure 4a,d illustrates the electron density for the bound drugs in two views related by a 180° rotation showing the bound magnesium ion (purple sphere) within the electron density envelope coordinated between the carbonyl at position 4 on the quinolone ring and the carboxyl at position 3. How this magnesium ion makes further but longer interactions, now of around 4 Å in length with the thymidine base at position 1 and the ParC D83 side chain, is also shown (figure 4b,c,e,f). This is probably coordinated through a water molecule whose density we do not see but is theoretically confirmed by placing waters in a hexacoordinated system around the Mg2+. These interactions are the same for both ACHN-245 and ACHN-454 (figure 4a–f). The carboxyl at quinolone position 3, besides interacting with this magnesium ion, also makes interactions with ParC S79 and R117; the lengths of these interactions are similar for both drugs. Thus, the drug molecules are in close register to ParC S79 and D83 residues, two mutational hotspots to quinolone resistance [33,34,53,54]. The amino group on the 7-(3-aminopyrrolidin-1-yl) group makes key interactions with a cluster of two glutamates and an arginine (ParE E474, E475 and R456 in ParE) through a single oxygen of the carboxyl groups of the glutamates and the main chain carbonyl of R456 (figure 4b,c,e,f). The hydrogen atoms of the cyclopropyl substituents are within a 4 Å range of the two of the oxygens of the phosphotyrosine Y118. Note that a second Mg2+ ion at each drug-binding site coordinated by ParE E433, D506 and D508 interacts with the DNA phosphodiester group between −1 and −2 (figure 4b,e) and, through repositioning in the absence of drug, may be involved in DNA strand breakage-reunion.
Figure 5 shows an edge-on view and illustrates how the ACHN-245 and 454 drugs are hemi-intercalated into a wedge-shaped binding pocket, forming π–π* stacking interactions between the aromatic quinolone rings and the DNA bases. In both cases, the cyclopropyl group on N-1 of the quinolone is cocked out of the plane of the quinolone. Figure 5 presents the key disposition of the Mg2+ ions, with the Mg2+ coordinated to the carbonyl and carboxyl of the drug together with S79 and D83 side chains, located behind as part of an alpha helix. At the top left of figure 5a,b, the second Mg2+ is coordinated to the catalytic cluster of E433, D506, D508, a glutamate/aspartate triad which has been found in many eukaryotic, archaeal and prokaryotic topoisomerases as well as in many recombinases, nucleases and polymerases [55–59]. In figure 6 are shown two orthogonal views of how the drugs are sandwiched between the bases superposed with our highest resolution (2.9 Å) structure of the clinafloxacin cleavage complex. The stacking interactions are very similar, but the cycloheptyl group of these Achaogen compounds projects towards R456 and D435 (figure 6a,c; figure 4c,f) and if modified could form a scaffold for substituents which could span towards these and other residues and impart further contacts. The cyclogroup of the Achaogen compounds projects towards the van der Waals cavity formed by the sugars of the DNA about the semi-intercalation and the only protein side chain making a close approach is E475 (figure 6a,c). In figure 6e,f are shown the superpositions of the Achaogen compounds with clinafloxacin in the complex 3RAD where the 7-group is free to rotate in contrast with its locked conformation in the Achaogen compounds.
The disposition of the quinolone drugs in the topoisomerase binding site is very similar to that of levofloxacin, and clinafloxacin, and the ring structure gives hindered rotation to the 7-(3-aminopyrrolidin-1-yl) moiety and maintains key interactions with E474, E475 and R456. The disposition of the cycloheptyl group shows how with further chemical elaboration of this scaffold, further interactions with neighbouring residues in the binding pocket may be formed.
Both antimicrobial susceptibility and enzyme-DNA cleavage assays reveal the 7,8-bridged fluoroquinolones ACHN-245 and ACHN-454 to be potent anti-pneumococcal agents. Table 5 presents the MIC data for the two compounds against quinolone-susceptible and quinolone-resistant S. pneumoniae strains. For comparison, data are included for levofloxacin, a widely used anti-pneumococcal quinolone, and for clinafloxacin, the experimental 8-chlorofluoroquinolone that bears the same 3-aminopyrrolidin-1-yl group at position 7 that is bridged to C-8 in the Achaogen compounds. Against the two quinolone-susceptible ATCC reference strains, the two ACHN compounds exhibited MICs of 0.03–0.06 mg l−1 which is comparable to or marginally better than clinafloxacin (0.06 mg l−1) and much lower than the MICs for levofloxacin, typically 0.5–1 mg l−1 (table 5). Clearly, chemical bridging of the 7 and 8 positions in the ACHN compounds does not compromise anti-pneumococcal activity.
Compared with wild-type strains, clinical isolates harbouring double quinolone-resistance mutations in gyrase and topoisomerase IV e.g. ParC79F and GyrA81F exhibited an 8–16-fold increase in MIC for ACHN-245 and ACHN-454 to 0.25–1 mg l−1, similar to the clinafloxacin MICs (table 5 and reported earlier in ). Thus, although quinolone-resistant strains have elevated MICs for ACHN-245 and ACHN-454, these MICs are very much lower than the typical levofloxacin MIC values of 32 mg l−1 (table 5). Evidently, as with clinafloxacin, the greater potency of the 7,8-bridged compounds extends to quinolone-resistant strains.
In complementary biochemical experiments, we examined drug-induced DNA cleavage mediated by purified wild-type S. pneumoniae topoisomerase IV and by its ParC79F and ParE475A mutants that exhibit resistance to quinolones and to antibacterial quinazolinediones, respectively (figures 7 and 8) . The trapping of a topoisomerase–DNA cleavage complex is thought to be the seminal event in quinolone and quinazolinedione action that underlies their cytotoxic action . In studies of the wild-type enzyme, we used the topo IV core domain—a ParE30-ParC55 fusion protein—that carries all the determinants necessary for DNA cleavage. The protein was incubated with supercoiled pBR322 DNA in the absence and presence of drug, and following addition of SDS and proteinase K, the DNA products were separated and analysed by agarose gel electrophoresis (figure 7). As expected, in the absence of drug, the core enzyme converted supercoiled pBR322 into a ladder of relaxed DNA topoisomers (figure 7, lanes a and b), a well-documented activity of this truncated topoisomerase IV complex not seen for the full-length holoenzyme (figure 8) . Inclusion of drug led to the formation of linear DNA in a dose-dependent fashion (figure 7). Comparing the CC25 values (the drug concentration that converted 25% of the supercoiled DNA substrate to the linear form), it is clear that the two 7,8-bridged quinolones exhibited a similar potency to clinafloxacin, with CC25 of 0.25–0.5 µM, about 10-fold more active than levofloxacin (figure 7; table 6).
Using full-length topoisomerase IV complexes with mutant ParC or ParE subunits, we could investigate the roles of ParC S79 and ParE E475 residues in drug action (figure 8). We could show that DNA cleavage mediated by the 7,8-bridged quinolones, levofloxacin and clinafloxacin was in each case much less efficient for the topoisomerase IV S79F mutant, with CC25 values 40–80-fold higher than seen with the wt enzyme (figure 8a; table 6). However, topo IV with a ParE 475A mutation showed only a twofold increase in CC25 for levofloxacin (reported previously in ), but a 10–20-fold increase for clinafloxacin, ACHN245 and ACHN454 (figure 8b; table 6). These results indicate that both ParC S79 and ParE E475 residues play a role in binding clinafloxacin and the 7,8-bridged quinolones (figures 4c,f and 6).
Previously, we showed that the dione-resistant ParE475A topoisomerase IV exhibited a similar 10–20-fold increase in the CC25 for PD 0305970 . We note that the ParE 475A mutation was selected with this quinazolinedione, which has a 3-aminopyrrolidin-1-yl group at position 7 closely similar to that present in clinafloxacin and the 7,8-bridged quinolones, but absent from levofloxacin. In conclusion, it may be that the presence of a 3-aminopyrroline side chain is a key factor in determining quinolone and dione interactions with ParE, and thereby the resistance profile. In any event, these biochemical experiments with mutant topoisomerase enzymes begin to map out potential drug–side chain interactions explored by the 7,8-bridged quinolones and by levofloxacin, a 1,8-bridged quinolone. It appears that the 7,8-bridging chemistry in the ACHN compounds does not reduce potency and suggests there is scope to explore further substitutions along the C1, C7 and C8 side of the quinolone that may enhance drug activity against resistant strains [61,62].
The ACHN-245 and ACHN-454 S. pneumoniae topoisomerase IV structures were deposited in the PDB with accession codes 4KPE and 4KPF, respectively.
We declare we have no competing interests.
This research was supported by the Biotechnology and Biological Sciences Research Council BBSRC UK (project grants BB/H00405X/1 and BB/K10069/1) in support of I.L. and X.-S.P. X.-S.P was also supported by St. George's, University of London. Research at Achaogen was funded in part by contract HDTRA1-07-C-0005 from the Defense Threat Reduction Agency, US Department of Defense.
We should like to thank the personnel of Diamond SLS, Oxfordshire, particularly from stations I02, I03, I04, I04–1 and I24 for their invaluable support and help with data collection. I.L., X.-S.P., D.A.V., R.T.C., A.W., H.E.M., L.M.F. and M.R.S. conceived and designed the experiments. I.L., D.A.V., X.-S.P. and M.R.S. performed the experiments. I.L., X.-S.P., D.A.V., L.M.F. and M.R.S. analysed and interpreted the data. X.-S.P., R.T.C., A.W., H.E.M. and L.M.F. contributed reagents/materials/analysis tools. I.L., X.-S.P., D.A.V., R.T.C., A.W., H.E.M., L.M.F. and M.R.S. wrote the paper and critiqued the output for intellectual content prior to publication.
- Received May 27, 2016.
- Accepted August 26, 2016.
- © 2016 The Authors.
Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.