We were interested in determining the basis of mycobacterial resistance to fosmidomycin. The target of fosmidomycin is the enzyme Dxr (DOXP reductoisomerase, Rv2870c), the first committed step in the non-mevalonate pathway of isoprenoid biosynthesis. M. tuberculosis Dxr has been expressed as a recombinant protein and is readily inhibited by fosmidomycin 1, indicating that resistance is not due to lack of a susceptible target. One possibility to account for the lack of whole cell sensitivity is that Dxr is not required for mycobacterial growth. The non-mevalonate pathway is the only known pathway of isoprenoid biosynthesis in mycobacteria 19, so it seems unlikely that this is the case. However, work using saturating transposon mutagenesis 20 gave rise to the prediction that dxr is not an essential gene.

In order to address this question, we used a two-step homologous recombination method to attempt to generate a defined mutant of dxr in M. tuberculosis 21. A deletion delivery vector, pORTAL3- with a defined deletion of dxr, was introduced into M. tuberculosis and single cross overs (SCOs) were obtained. Double cross-overs (DCOs) were then generated from the SCOs and screened by PCR to determine if the wild type or the deletion allele was present. Of the 40 DCOs screened all were found to be wild type, suggesting essentiality of dxr.

Since we were unable to isolate a deletion mutant in the wild type background, we constructed a merodiploid strain carrying the entire dxr operon with its native promoter on a mycobacteriophage L5 based integrating vector (pOPW: dxr^intgm). DCOs were generated in this background as before and screened by PCR. Of the 24 DCOs screened 12 had the deletion allele (50%), indicating that we were able to isolate chromosomal deletion of dxr in this background. The expected genotype of a number of these strains, which we have termed "

del

in

t

Since dxr is in an operon (Figure 1A), it is possible that the inability to isolate mutants could be due to one of the downstream genes being disrupted in the mutant. Analysis of the mutation that would be created by the pORTAL3 delivery vector suggests that it may disrupt the promoter region at the start of the operon and also the gene downstream of dxr (Rv2869c), a non-essential intramembrane-cleaving protease (iCLIP) 2223.

In order to confirm that dxr alone is essential, we used a gene switching strategy. This relies on the fact that efficient replacement of a resident integrated vector can easily be achieved in M. tuberculosis by transformation and selection for a second integrating plasmid carrying a different antibiotic resistance marker 242526. We made use of the delinquent strain carrying the complete operon (pOPW) (dxrΔ, dxr^int gm) and tested the ability of a plasmid carrying the whole operon except dxr (pOPD; Figure 1A) to complement the chromosomal mutation. The delinquent strain was transformed with pOPD and plated onto hygromycin plates to select for the incoming vector. No hygromycin resistant transformants were obtained, although a control vector transformed the same cells at an efficiency of 1.4 × 10⁶per μg. In contrast replacement of pOPW by pOPD in a wild-type background was easily achieved at a efficiency of 1.2 × 10⁶ per μg. These data indicate that a vector lacking only dxr does not complement the chromosomal mutation, and therefore that dxr itself is essential. It is still possible that other genes in the operon are essential as well and this strategy could in the future be extended to examining the essentiality of other genes in this operon.

We confirmed that mycobacteria were highly resistant to fosmidomycin resistance in our culture conditions. We looked at the growth of M. smegmatis and M. tuberculosis in liquid medium in the presence of fosmidomycin. No inhibition of growth was seen at any concentration up to 1 mg/ml (Figure 2A & B), confirming that both species are resistant to a high level. Since dxr is essential in normal culture conditions and that Dxr is sensitive to fosmidomycin inhibition 1, there must be another explanation for the high level resistance.

Resistance to fosmidomycin has been attributed to the Fsr efflux pump in several micro-organisms 1415. Mycobacteria are known to posses many efflux pumps which confer intrinsic antibiotic resistance 272829303132. We investigated whether efflux plays a role in fosmidomycin resistance in mycobacteria, by utilising the efflux pump inhibitors, reserpine, verapamil and carbonyl cyanide m-chlorophenylhydrazone (CCCP). The concentrations of efflux pump inhibitors we used were taken from a previous study in M. smegmatis 27. As before, reserpine and verapamil at 12 and 40 μg/ml respectively were not inhibitory to growth. However, CCCP at a concentration of 15 μg/ml was seen to significantly inhibit growth in both M. smegmatis and M. tuberculosis. Therefore, we tested lower concentrations of CCCP; 1.25 μg/ml was the highest permitted concentration where growth rate was not notably affected, although some inhibition was still seen.

Growth of mycobacteria in the presence of efflux pump inhibitors and fosmidomycin was assayed. All cultures grew to an OD of above 1.0 (approx 10⁹ cell per ml), indicating that there were no significant effects on growth rate or the final OD reached. No difference in the sensitivity of either species to fosmidomycin was seen in the presence of the inhibitors, indicating that efflux is not the mechanism of resistance.

Since we had discounted efflux as the resistance mechanism, we determined whether fosmidomycin was transported into the mycobacterial cells in the first place. Uptake of antibiotics is typically measured using a radiolabelled derivative. However, radio-labelled fosmidomycin is not available commercially and there are additional associated hazards connected with using radio-isotopes under Category Three level containment in the UK. A microbiological bioassay, based on the inhibition of E. coli by active compounds, has previously been used to measure fosmidomycin concentrations 1133. In this assay, active fosmidomycin is measured by inhibition of E. coli growth on disks. A standard curve using known concentrations of fosmidomycin can be generated and used to quantify fosmidomycin [see Additional file 1] 34.

M. smegmatis, M. tuberculosis and E. coli cells were incubated with 1 mg/ml of fosmidomycin for 30 min (Figure 3A). Active fosmidomycin in the supernatant (extra-cellular) and cell-free extracts (intra-cellular) was measured. No active fosmidomycin was present in the cell-free extracts generated from M. tuberculosis or M. smegmatis; all of the fosmidomycin remained in the supernatant. In contrast, E. coli extracts contained more than 150 μg/ml and a reduction in the amount of fosmidomycin in the supernatant was seen. These data confirmed fosmidomycin uptake in E. coli, and conversely that there is a lack of uptake by mycobacteria.

The limit of detection for uptake in this assay was 0.1 μg/ml, which is 10,000-fold less than the level of antibiotic in the medium. One of the limitations of using a high concentration of fosmidomycin with E. coli is that some of the antibiotic in the supernatant could have been released from the cells after antibiotic-induced lysis. However, this would lead us to underestimate uptake, rather than overestimate it.

The results from the transport assay strongly suggested that fosmidomycin resistance in mycobacteria results from a lack of active transport. However, it was not possible to tell if the antibiotic was being modified in the previous assay. This was because a high concentration of antibiotic was used; if only a small proportion of the fosmidomycin were modified, it would not have been detected as a loss of fosmidomycin activity in the supernatant. Thus, the possibility that the cells might be modifying the antibiotic into a non-toxic form still existed. To address this question we repeated the exposure assay using a lower concentration of fosmidomycin which allowed more precise quantification of the active antibiotic in the supernatant. M. smegmatis or E. coli cells were incubated with 20 μg/ml of fosmidomycin for 60 min and active fosmidomycin assayed in the supernatant (Figure 3B). In the supernatant exposed to M. smegmatis, a 2-fold reduction in activity (from 20 to 10 μg/ml) was seen. Given the limitation of the assay, where such a difference would result in a change in the zone of inhibition of less than 1 mm, this is likely to be due to lack of sensitivity in the assay. In addition, if detoxification were occurring we would expect to see a complete reduction of fosmidomycin activity. In contrast a 100-fold reduction in fosmidomycin activity was seen in the E. coli supernatants (due to uptake from the extra-cellular environment into the cell). Thus, no modification of the antibiotic by M. smegmatis was seen.

Although fosmidomycin is not taken up into the mycobacterial cells, there could still be intra-cellular enzymes capable of detoxification. In order to address this, we determined whether extracts of M. smegmatis were capable of inactivating fosmidomycin. M. smegmatis cell-free extracts were incubated with fosmidomycin (20 μg/ml and 200 μg/ml) for 10 to 60 minutes. No decrease in fosmidomycin activity was seen during this time, indicating that no detoxifying enzymes are present (Figure 4). These data confirmed that resistance is not due to modification of fosmidomycin.

M. tuberculosis H37Rv was cultured in Middlebrook 7H9 liquid medium supplemented with 10% v/v OADC (oleic acid, bovine serum albumin, D-glucose, catalase; Becton Dickinson) and 0.05 % w/v Tween 80 (7H9/Tw/OADC) or on solid Middlebrook 7H10 agar supplemented with 10% v/v OADC (7H10/OADC). M. smegmatis mc²155 was grown on Lemco media 39. X-Gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) was used at 50 μg/ml; IPTG (isopropyl-beta-D-thiogalactopyranoside) at 0.5 mM; kanamycin at 20 μg/ml; hygromycin B at 100 μg/ml; and sucrose at 2% w/v. Fosmidomycin (sodium salt) was purchased from Invitrogen. Efflux pump inhibitors: reserpine was used at 12 μg/ml; verapamil at 40 μg/ml; and carbonyl cyanide m-chlorophenylhydrazone (CCCP) at 1.25 μg/ml. M. tuberculosis growth curves were conducted in 4 ml of 7H9/Tw/OADC media in 16 mm glass tubes containing an 8 mm magnetic stirrer bar, stirring at 150 rpm.

A deletion delivery vector for dxr (pORTAL3) was constructed as follows (Figure 1): the upstream flanking region was amplified from M. tuberculosis genomic DNA using the primer pair DXR US F (PstI) 5'-TGGG

CTGCAG

GCGGCCGC

GCGGCCGC

AAGCTT

A complementing vector carrying the entire dxr operon with its native promoter was constructed by amplifying a 4.7 kb region from genomic H37Rv DNA with primers DXR OP C' F 5'-TTAATTAACACCTCGCGGTACAGCTCGTC-3' and DXR OP C' Rev 5'-TTAATTAAGTCGTCAGCGCTGATATGCCC-3'. The product was cloned into pSC-A (Stratagene); to give dxr_op/pSCA, and the Gm-int cassette from pUC-Gm-Int 41 was introduced as a HindIII fragment to give pOPW (Figure 1).

We constructed a merodiploid strain by electroporating the dxr SCO with pOPW and isolating kanamycin/hygromycin/gentamicin resistant transformants. DCOs were isolated and PCR-screened as before. Strains were confirmed by Southern analysis using the AlkPhos Direct system (GE Healthcare) to label and detect DNA. A deletion derivative of dxr_op/pSCA was constructed in which the 1.2 kb dxr gene was removed by site directed mutagenesis with primers DXR KO SDM F 5'-GTGACCAACTCGACCGACGTATCTGGTATGGCTTCG-3' and DXR KO SDM Rev 5'-CGAAGCCATACCAGATACGTCGGTCGAGTTGGTCAC-3'. The Hyg-Int cassette from pUC-Hyg-Int 41 was cloned in via a HindIII digest. The resulting plasmid, pOPD (Figure 1), was used in a gene switching experiment 25. Strains were electroporated with integrating vectors and transformants isolated on the appropriate antibiotic selection (for the incoming vector). Transformants were patch tested for antibiotic resistance as required.

Bacteria were grown to late log phase (OD₆₀₀ = 1.0, approximately 10⁹ cells per ml), pelleted by centrifugation at 3 000 × g for 5 minutes and re-suspended in 50 mM Tris buffer pH 7.0 at approximately 10¹¹ cells per ml. One ml of suspension was added to 0.5 ml fosmidomycin solution and 3.5 ml LB media for E. coli and M. smegmatis or Middlebrook 7H9/Tw/OADC medium for M. tuberculosis and incubated at 37°C for 30 or 60 min with shaking at 100 rpm (E. coli, M. smegmatis) or standing (M. tuberculosis). Cells were pelleted by centrifugation at 3 000 × g for 5 minutes, the supernatant was collected, filter sterilized through a 0.2 μm syringe filter and stored on ice. Cells were washed twice with 5 ml of 50 mM Tris pH 7.0, re-suspended in 1 ml of 10 mM Tris pH 7.0, transferred to a Fast Prep Lysing Matrix B tube (QBiogene) and incubated on ice for 5 min. Cell-free extracts were prepared using the Fast Prep at speed 4 for 30 seconds (QBiogene). Samples were incubated on ice for 5 min, cell debris pelleted, the cleared lysate recovered, filter sterilized and stored on ice.

Fosmidomycin was measured using a microbiological assay 34. 100 ml of cell-free extracts or supernatants (and serial dilutions) were applied to a sterile 14 mm filter paper disc (Whatman). The discs were allowed to dry at room temperature for 15 minutes before transferral onto individual 90 mm agar plates containing 0.5% v/v of an E. coli overnight culture. Plates were incubated for 20 hours at 37°C and the diameter of the zone of bacterial growth inhibition was measured. Discs with known concentrations of fosmidomycin were used to make a standard curve, to calculate the concentration of fosmidomycin in the test solutions. The concentration of fosmidomycin in the samples was then calculated from the standard curve.