A novel method to determine antibiotic sensitivity in Bdellovibrio bacteriovorus reveals a DHFR-dependent natural trimethoprim resistance

Bdellovibrio bacteriovorus is a small Gram-negative bacterium and an obligate predator of other Gram-negative bacteria. Prey resistance to B. bacteriovorus attack is rare and transient. This consideration together with its safety and low immunogenicity makes B. bacteriovorus a valid alternative to antibiotics, especially in the treatment of multidrug resistant pathogens. In this study we developed a novel technique to estimate B. bacteriovorus sensitivity against antibiotics in order to make feasible the development and testing of co-therapies with antibiotics that would increase its antimicrobial efficacy and at the same time reduce the development of drug resistance. Results from tests performed with this technique show that among all tested antibiotics, trimethoprim has the lowest antimicrobial effect on B. bacteriovorus. Additional experiments revealed that the mechanism of trimethoprim resistance in B. bacteriovorus depends on the low affinity of this compound for the B. bacteriovorus dihydrofolate reductase (Bd DHFR).

a consequence of this phenomenon, B. bacteriovorus predation while killing a large percentage of prey, in vitro, never results in complete eradication of the attacked population 11,12 , regardless of the initial predator to prey ratio (PPR) in the bacterial pool.
One possibility is to use B. bacteriovorus in combination with conventional antibiotics to potentiate the effectiveness of the antibiotic therapy against Gram-negative pathogens (including MDR cells within the population). In order to make this approach feasible more has to be known about the ability of B. bacteriovorus to resist antibiotics itself. Standard assays to determine the Minimum Inhibitory Concentration (MIC) are not suitable for B. bacteriovorus, since for growth, this species requires the presence of bacterial prey cells (which are themselves susceptible to antibiotics). The use of B. bacteriovorus in a co-therapy with antibiotics has already been proposed 9,13 . In a study by Im et al. 14 , B. bacteriovorus has been used successfully in combination with violacein, an antibiotic effective against Gram-positive bacteria, to treat MDR pathogens. The combination of both antibacterial agents allowed to expand their spectrum of action (that in case of B. bacteriovorus is restricted to Gram-negative bacteria) to treat Gram-positive and -negative mixed bacteria populations.
In this study, we developed a novel assay in liquid medium, which allows testing of B. bacteriovorus sensitivity against antibiotics. For this purpose, we tested antibiotics belonging to different classes of pharmaceuticals with different specificities for Gram-positive and -negative bacteria. B. bacteriovorus growth is estimated by the reduction in absorbance at 600 nm (caused by the lysis of prey cells) of mixed B. bacteriovorus/E. coli cultures in the presence and absence of different antibiotics concentrations. Antibiotic minimal inhibitory concentration (MIC) values are subsequently determined by non-linear regression. Our findings show that among all tested antibiotics, trimethoprim (TMP) is the least active on B. bacteriovorus.
The most common mechanism of natural resistance against TMP is caused by the lack of affinity of the dihydrofolate reductase (DHFR) for this antibiotic compound. In order to ascertain whether this mechanism is the cause of resistance in B. bacteriovorus, we cloned three B. bacteriovorus genes (bd3231, bd0323, bd1356) with homologies to the E. coli folA chromosomal gene (b0048, encoding the DHFR enzyme) and analyzed their sensitivity against TMP by plate dilution assays. By this screening we identified the B. bacteriovorus DHFR (Bd DHFR) as the protein encoded by the bd3231 gene. The involvement of certain amino acids (V6, M29 and F51) in TMP binding is also investigated by site-directed mutagenesis, plate dilution and MIC assays. The kinetic properties and TMP inhibition of activities are studied on the purified protein and its substituted variants.

Results
Determination of B. bacteriovorus antibiotic sensitivity. B. bacteriovorus is a relatively small-sized bacterium (0.25-0.35 by 0.5-2.5 μm) and therefore does not scatter light efficiently at 600 nm. Direct cell growth measurements of B. bacteriovorus upon prey bacteria can be estimated by reduction of the optical density at 600 nm (OD 600 ) (i.e. the loss of culture turbidity by B. bacteriovorus-induced lysis of prey cells). In this study, we designed a 96-well based method for the determination of antibiotic MICs for B. bacteriovorus. B. bacteriovorus HD100 cells were added to an E. coli culture and the reduction of OD 600 measured in absence and presence of different concentrations of antibiotics. Assays were conducted in a solution of Ca/HEPES buffer and the exhausted medium transferred with the E. coli overnight culture and the B. bacteriovorus dilution. Therefore, E. coli cells were in the stationary phase and their growth during the assays was negligible. In Fig. 2a is an example of the MIC value determination for TMP (results from a single PPR series). MIC values are dependent on the initial PPR used in the assay, as standard MIC assays are dependent on the initial bacterial concentration. However, B. bacteriovorus numbers and therefore, the initial PPR can be estimated reliably only after the assay by plating in double layer agar plates. The relation between initial PPR and MIC values was then investigated by adding different numbers of B. bacteriovorus cells to the fixed number of E. coli prey cells in parallel replicated experiments (on different rows of the same plate containing replicated antibiotic serial dilutions). We also considered the effect of antibiotics on the E. coli prey cells and measured the OD 600 values of E. coli in the absence of B. bacteriovorus (PPR = 0), at the same antibiotic concentrations. B. bacteriovorus growth is estimated by subtracting absorbance values of B. bacteriovorus at different PPRs from the values of E. coli (PPR = 0) at the same antibiotic concentrations. Therefore, for each PPR series of data is extrapolated a Δ curve of B. bacteriovorus growth, in which the 100% growth refers to absorption values of samples not treated with antibiotic (corresponding to the highest level of prey lysis that is achieved at that particular initial PPR). The efficacy of an antibiotic on B. bacteriovorus is displayed by the lack of E. coli lysis. In order to determine the MIC of an antibiotic, B. bacteriovorus growth data at different PPRs (plotted against the logarithm of antibiotic concentration) were fitted with a modified Gompertz equation 15 (Fig. 2b). Antibiotic MIC values were extrapolated from each curve as the minimal antibiotic concentration that prevents B. bacteriovorus growth. The theoretical MIC value was calculated for all antibiotics by linear regression at a PPR of 0.02 in a plot of MIC values against the logarithm of initial PPRs (Fig. 2c) Table S1). However, TMP gained our attention as it is the antibiotic characterized by the highest MIC for B. bacteriovorus with a value equal to 223 ± 24 µg/mL. Bd3231 DHFR confers TMP resistance to E. coli. The known target of TMP is the dihydrofolate reductase (DHFR), an enzyme that catalyzes the reduction of 7,8-dihydrofolate (DHF) to 5,6,7,8-tetrahydrofolate (THF) by using nicotinamide adenine dinucleotide phosphate (NADPH) as cofactor [16][17][18][19] . This reaction is essential for the de novo synthesis of purines and some amino acids. A common mechanism of resistance against TMP is the lack of affinity of this compound for DHFR. To test whether the B. bacteriovorus DHFR (Bd DHFR) is insensitive towards TMP, we cloned three folA homologs from B. bacteriovorus (bd3231, bd0323, bd1356) in E. coli and analyzed them by plate dilution assays (Fig. 3). The B. bacteriovorus gene bd3231 confers very high resistance against TMP, even exceeding the resistance caused by the recombinant E. coli folA, despite its apparent two-fold higher expression levels (Figs. S1 and S2). The other B. bacteriovorus genes were expressed at levels of apparent 1.3-fold the E. coli folA expression, however, did not confer a TMP-resistant phenotype. Thus, we named Bd3231 as Bb DHFR.
Role of selected amino acids in tMp binding to Bd DHfR. We deduced from a DHFR amino acid sequence alignment of TMP sensitive vs. resistant bacteria, in combination with the structural insights from binding of TMP to the E. coli DHFR 20 (Ec DHFR; Fig. S3, Table S2) and Coxiella burnetii DHFR (PDB: 3TQ8 21 ; Fig. 4), that three residues (V6, M29, and F51) might be responsible for the TMP-resistance observed for the Bd DHFR. To further analyze this hypothesis, we single-substituted these three residues in the Bd DHFR with the residues present in the TMP-sensitive DHFRs. These mutant genes were subsequently expressed in E. coli and tested for The M29L and F51I substitutions confer a lower level of resistance, a strong indication that these amino acid residues could be involved in the TMP resistance mechanism. However, the higher expression of the M29L and V6I_M29L DHFR substitution variants might be an alternative explanation for the observed correlated resistance pattern towards TMP (Fig. S4). In order to analyze the effects of the substitutions on TMP binding independent of the expression levels, we overproduced and purified the wild type and substituted DHFR variants and conducted kinetic and TMP inhibition measurements.
DHfR in vitro assays. In this study we developed a new plasmid, pBXKCPD, which is compatible with the FX-cloning method 22 and allows to express target proteins with a His 10 -tagged CPD 23 as C-terminal fusion (Fig. S6). This system allowed us to optimize and enhance cloning, expression and purification procedures. The Bd DHFR wildtype, the substituted variants and the Ec DHFR were overproduced and purified by Ni-NTA affinity chromatography and their purification tag removed by auto-catalytic cleavage (the Bd DHFR wild type was further subjected to size-exclusion chromatography).  www.nature.com/scientificreports www.nature.com/scientificreports/ The kinetic analysis of the different DHFR variants (Bd DHFR wild type, Ec DHFR, Bd DHFR_M29L, Bd DHFR_F51I and Bd DHFR_M29L_F51I) shows that at high substrate concentrations, a reduction in the initial velocity is apparent and describes a mechanism known as substrate inhibition (Fig. S7). In the Ec DHFR at high substrate concentrations, this mechanism involves the formation of the two dead-end complexes DHFR-DHF-NADP + and DHFR-DHF-THF 16 .
Results from TMP inhibition assays (Table 2, Fig. S8) show that the Bd DHFR, with a K i of 49 ± 15 nM, is 2,450 times more resistant against TMP inhibition than Ec DHFR (K i of 0.02 nM 24,25 ). Furthermore, it is evident that both the M29L and F51I substitutions cause a decrease in affinity for TMP, leading to a more inhibitor-sensitive enzyme. This is particularly evident in the double mutant, Bd DHFR_M29L_F51I, which has a K i equal to 3.2 ± 1.6 nM, over 15 times lower than the wild type variant. The results from this analysis corroborates the results obtained from agar plate dilution and MIC experiments of E. coli expressing DHFR substituted variants. In light of this, we can conclude that the TMP sensitive phenotype observed in E. coli expressing M29L and F51I DHFR substituted variants is due to the enhanced TMP affinity of these DHFRs.

Discussion
A novel method to determine antibiotic sensitivity in Bdellovibrio bacteriovorus. The method developed in this study allows to determine the degree of resistance of B. bacteriovorus against antibiotics relative to a reference organism (in this study E. coli BW25113). MIC values for antibiotics with B. bacteriovorus are prone to variation depending on the reference organism and growth conditions (i.e. culture broth, initial PPR, temperature, aeration and shaking force) 26 . The MIC values for both organisms were determined by fitting percentage growth values with a modified Gompertz equation 15  This MIC method can be modified in order to test B. bacteriovorus against other pathogens and/or other antibiotics by changing the culture broth or the vessel to increase or reduce aeration according to requirements (e.g. 96-well plates rather than tubes or flasks). The main limitations of this technique are the use of non-standard conditions such as a culture broth containing Ca 2+ and Mg 2+ ions and a reduced temperature, which are required for B. bacteriovorus cultivation. Furthermore, this method may not allow for the determination of absolute MIC values of bactericidal antibiotics at concentrations equal or higher than their minimal bactericidal concentration (MBC). B. bacteriovorus is not able to replicate on heat-inactivated prey bacteria (with the exception of B.   www.nature.com/scientificreports www.nature.com/scientificreports/ bacteriovorus host-independent strains 27,28 ), while it can grow on UV-inactivated cells 29 . To the best of our knowledge, there are no studies regarding B. bacteriovorus ability to grow on antibiotic-inactivated cells. Therefore, prey bacteria inactivation at concentrations equal or above the MBC may prevent B. bacteriovorus from growing, resulting in a lower estimate of the MIC for B. bacteriovorus (this is true especially for lysed preys or preys with a compromised structural integrity).
The MBC values of bactericidal antibiotics are usually equal to, and generally do not exceed, four-fold the MIC values. On the other hand, the MBC values of bacteriostatic antibiotics are usually many folds over the related MIC values. Bacteriostatic antibiotics include macrolides, tetracyclines, sulfonamides, trimethoprim, chloramphenicol and linezolid, while bactericidal antibiotics include β-lactams, aminoglycosides, fluoroquinolones and vancomycin. Considering that MBC values are higher than the related MIC values, the determination of the MIC values for B. bacteriovorus may be underestimated only for those bactericidal antibiotics characterized by a MIC ratio (Bd/Ec; Table 1) above or equal to 1. However, whether prey bacteria are lysed or not is irrelevant for the determination of the Δ curve of B. bacteriovorus growth. Indeed, this curve is obtained by subtracting OD 600 values of test samples (B. bacteriovorus at different PPRs) from OD 600 values of E. coli control samples (PPR = 0) measured at the same antibiotic concentrations. Therefore, an eventual decrease in the OD 600 caused by the activity of a bactericidal antibiotic will be measured in both control and test samples and will cancel out the difference between the two curves.
Bdellovibrio bacteriovorus DHfR-dependent natural trimethoprim resistance. Our findings demonstrate that the B. bacteriovorus mechanism of resistance against TMP is caused by the low affinity of the DHFR for this compound, which is dependent on the presence of primarily two amino acid residues (i.e. M29 and F51) in the binding site. The Bd DHFR structure prediction analysis performed with Phyre 2 30 using the C. burnetii DHFR (PDB entry: 3TQ8 21 ; Fig. 4) as template suggests that M29 and particularly F51 may be in steric clash with TMP and/or reduce the cavity volume of the catalytic site disfavoring TMP binding. This hypothesis is supported by the substitution-induced increase in affinity of Bd DHFR for TMP by M29L and F51I. Both introduced side-chains are characterized by a reduced steric hindrance in comparison to the substituted residues, while their net charge and hydrophobicity are virtually unchanged.
Although the MIC method for B. bacteriovorus described in this work employs non-standard conditions, it allows for the first time to detect and compare B. bacteriovorus antibiotic resistance levels with other bacteria. In the perspective of employing B. bacteriovorus for co-therapies with antibiotics, its antibiotic resistance could be increased by insertion in the genome of resistance genes. This acquired resistance could be further investigated by the technique developed in this study and would allow the use of B. bacteriovorus in combination with further antibiotics. This study might pave the way for testing B. bacteriovorus (and other predatory bacteria) based co-therapies with antibiotics in both mammalian cell cultures and animals.

Materials and Methods
construction of bacteria strains. All bacteria strains used in this study are listed in Table S3 of supporting information. Plasmids (Table S4) were constructed and genes cloned and/or modified as explained in the supporting information by using the primers listed in Table S5. B. bacteriovorus antibiotic resistance determination. B. bacteriovorus HD100 predatory growth was synchronized by sub-culturing into a YPSC (Table S1) E. coli BW25113 prey overnight culture (using a 2% inoculum), every 24 hours for 3 days as explained elsewhere 31 .
Two-fold serial dilutions of antibiotics (Table S6) were made along the rows of 96-well plates (BRANDplates pureGrade TM S, no. 781662) in 100 µL Ca/HEPES buffer (5.94 g/L HEPES free acid, 0.284 g/L calcium chloride 2·H 2 O, adjusted to pH 7.6 with NaOH), including controls (no antibiotic present). B. bacteriovorus cultures were diluted by a factor of 25 in Ca/HEPES buffer. B. bacteriovorus two-fold serial dilutions were prepared in a new sterile 96-well plate along the column line (using as diluent a suspension of heat-inactivated B. bacteriovorus), including controls (no viable B. bacteriovorus present). Fifty µL of these dilution series were transferred from each well of the latter plate to the corresponding wells of plates with antibiotic dilution series. Subsequently, 50 µL of an E. coli BW25113 overnight culture grown in YPSC broth was added to each plate well. At this point, which corresponds to time zero, the blank was determined by measuring the OD 600 of cell suspensions in a plate reader (TECAN Infinite 200 Pro). Final OD 600 measurements were carried out after an incubation of 24 hours at 29 °C under shaking conditions (300 rpm, 3 mm throw). The number of E. coli cells was estimated by reading the OD 600 of the overnight culture prior to the beginning of the assay. The number of viable B. bacteriovorus cells was estimated by plating in double layer YPSC agar plates as explained elsewhere 31 . Following an incubation time of three days at 29 °C, the number of B. bacteriovorus cells in the initial inoculation culture was estimated by counting plaque forming units (PFU). Enumeration of B. bacteriovorus by plaques count ensures that only viable cells are counted and, in our experience, is preferable over other methods based on microscopy and/or fluorescence. Pictures of plaques were acquired by the aid of a plaque visualization chamber, specially developed in this study for this purpose (Fig. S9).
All absorbance values were subtracted of the related blanks. Then, absorbance values of test samples (B. bacteriovorus co-cultures) were subtracted from the control values (E. coli without B. bacteriovorus) and plotted in graph against the logarithm of antibiotic concentrations. The resulting Δ curve is an indication of B. bacteriovorus growth. Data was expressed as percentage of growth and plotted in a graph of log(inhibitor) vs. response. The minimal inhibitory concentration (MIC) value was defined as the concentration of antibiotic that prevent B. bacteriovorus growth and were calculated by fitting B. bacteriovorus growth inhibition curves with a Gompertz modified equation 15  www.nature.com/scientificreports www.nature.com/scientificreports/ E. coli antibiotic resistance determination. Serial antibiotic dilutions in 100 µL Ca/HEPES were made along the rows of 96-well plates including controls without antibiotic. E. coli BW25113 cells from an overnight culture were washed and re-suspended in Ca/HEPES buffer to a final OD 600 of 0.024. Aliquots of 50 µL of this E. coli suspension and 50 µL of a fresh YPSC broth were added to the wells of these plates and the blank measured at 600 nm. Final OD 600 measurements were carried out after an incubation of 24 hours at 29 °C under shaking conditions. Data is shown as percentage of growth and plotted in a graph against the logarithm of antibiotic concentration. MIC values were calculated by fitting data with a Gompertz modified equation 15 (using GraphPad Prism 5).
Agar plate dilution assays. E. coli BW25113ΔacrB and E. coli MC1061 were transformed within one week prior to the assay, with either vectors derived from p7XC3H and pBXKCPD bearing folA variants, respectively. Overnight cultures were grown in LB broth supplemented with the appropriate antibiotic marker and then diluted to the OD 600 of 1 with the fresh broth. Five µL of two-fold serial dilutions of these cultures were spotted onto 1.5% (w/v) agar Müller-Hinton II (MHII; Roth) plates supplemented with the antibiotic marker and TMP at different concentrations. The plates used to test E. coli BW25113ΔacrB bearing p7XC3H derived vectors were supplemented with 0.01 mM IPTG, while the plates used to test E. coli MC1061 bearing pBXKCPD derived vectors were supplemented with 0.05% ( All purification steps of overexpressed proteins explained below were carried out at 4 °C and using filter-sterile (0.2 µm), ice-cold solutions. Cells were pelleted at 15,000 g and washed in cell wash buffer (300 mM NaCl, 20 mM Tris-HCl, pH 8.5 (4 °C), 10 mM imidazole pH 8.5 (4 °C)). Cell pellets were re-suspended in 3 mL lysis buffer (0.1% Triton X-100, 1 mM β-mercaptoethanol, 300 mM NaCl, 20 mM Tris-HCl, pH 8.5 (4 °C), 10 mM imidazole pH 8.5 (4 °C), 10% (v/v) glycerol, 1 mM MgCl 2 , trace amounts of DNAse I, 1 mM PMSF, 1 mg/ml lysozyme) per g of wet weight and incubated on ice for 30 minutes. Cells were lysed by sonication and the lysates cleared by centrifugation at 6,000 g for 20 minutes. Cell debris was removed by two low speed centrifugation steps (20 minutes at 6,000 g and 15 minutes at 15,000 g). Supernatant was applied onto a Ni-NTA Sepharose column (Qiagen). Subsequently, the flow-through was collected and the column washed with wash buffer (0.1% Triton X-100, 1 mM β-mercaptoethanol, 300 mM NaCl, 20 mM Tris-HCl pH 8.5 (4 °C), 30 mM imidazole pH 8.5 (4 °C) and 10% (v/v) glycerol). All DHFR variants were eluted from columns by the addition of 10 mL elution buffer (0.1% (v/v) Triton X-100, 1 mM β-mercaptoethanol, 300 mM NaCl, 20 mM Tris-HCl pH 8.5 (4 °C), 10 mM imidazole pH 8.5 (4 °C), 10% (v/v) glycerol, 5 µM NADPH and 200 µM inositol hexakisphosphate) per bed volume after an overnight incubation (>16 hours) in batch with mixing by rotary inversion. Eluates were concentrated using Amicon Ultra centrifugal filters with a cut-off of 10 kDa. The concentrated solutions were centrifuged for 20 minutes at 16,000 g to remove insoluble aggregates, after which the supernatant was flash frozen in liquid nitrogen and stored at −80 °C until use.
DHfR in vitro assays. DHFR activities were determined in a plate reader (TECAN Infinite 200 Pro) by measuring the decrease in absorbance at 340 nm at 25 °C. Measurements were performed as triplicates in clear 96-well plates (BRANDplates pureGrade TM S, no. 781662) every 15 s, over a time of 10 minutes in the presence of assay buffer (20 mM Tris-HCl, 150 mM NaCl, 0.1 mM EDTA, 1 mM β-mercaptoethanol, pH 7.5 at 25 °C). The conversion of substrate into product was quantified using an extinction coefficient for the reaction of 12.3 mM −1 cm −1 32 . The concentrations of DHF and NADPH were determined spectroscopically using the extinction coefficients of 28,000 M −1 cm −1 at 282 nm and 6,200 M −1 cm −1 at 339 nm, respectively. The concentration of active enzymes was determined by titration with methotrexate 33 . The hysteresis effect (slow change in reaction velocity) was prevented by concentrating the DHFR enzymes in the master mix solutions 10 times their working concentrations 34 .
The kinetic values K m (the Michaelis-Menten constant relative to the substrate, DHF) and V max , for all DHFR enzymes, were determined via kinetic assays in which the concentration of co-substrate, NADPH, was kept constant in all reactions at a saturating level while the substrate, DHF, was tested at different concentrations. Initial www.nature.com/scientificreports www.nature.com/scientificreports/ velocities were plotted in a graph against substrate concentrations and fitted with the "Substrate inhibition" equation (GraphPad Prism 5; Fig. S7