To test the above hypothesis, we examined mutation rates in E. coli
strain MG1655 following treatment with low levels of the bactericidal antibiotics, norfloxacin (quinolone), ampicillin (β-lactam) and kanamycin (aminoglycoside), respectively. Mutation rates were determined by plating aliquots of treated cultures onto rifampicin plates, counting rifampicin-resistant colonies, and using the MSS maximum likelihood method (Rosche and Foster, 2000
) to estimate the number of mutation events per culture (see Experimental Procedures for additional details). The mutation rate for untreated wildtype E. coli
was approximately 1.5×10−8
Treatment with 1µg/ml ampicillin, 3µg/ml kanamycin, 15ng/ml norfloxacin, or 50ng/ml norfloxacin resulted in significant increases in the mutation rate relative to an untreated control (). Treatment with 1µg/ml kanamycin resulted in a modest increase in mutation rate (). The largest increases in mutation rate were seen following treatment with ampicillin or 50ng/ml norfloxacin (). These changes were on par with the increase in mutation rate observed following treatment with 1mM hydrogen peroxide (), a concentration of hydrogen peroxide known to induce hydroxyl radical formation via Fenton chemistry (Imlay et al., 1988
). To determine if there is a correlation between these changes in mutation rate and ROS formation, we examined radical levels using the radical-sensitive dye hydroxyphenyl fluorescein, HPF (Setsukinai et al., 2003
) (see Experimental Procedures for more details). We found a significant correlation (R2
= 0.8455) between the fold change in mutation rate and peak HPF signal for the treatments described above ().
Low levels of bactericidal antibiotics increase mutation rate due to reactive oxygen species formation
The strong correlation between ROS formation and fold change in mutation rate following treatment with bactericidal antibiotics suggests that ROS actively contribute to bactericidal drug-induced mutagenesis. To test if this is indeed the case, we added 100mM thiourea to wildtype E. coli
treated with antibiotics or hydrogen peroxide at the concentrations noted above (). Thiourea is a potent hydroxyl radical scavenger which mitigates the effects of hydroxyl radical damage in both prokaryotes and eukaryotes (Novogrodsky et al., 1982
; Repine et al., 1981
; Touati et al., 1995
). We have previously shown that thiourea reduces hydroxyl radical formation and cell killing following treatment with bactericidal antibiotics (Kohanski et al., 2007
The addition of thiourea significantly reduced the mutation rate to near untreated levels following the addition of 1mM hydrogen peroxide, norfloxacin, or ampicillin (). Interestingly, we were unable to detect any rifampicin-resistant colonies after plating up to 109 cells following treatment with both 3µg/ml kanamycin and thiourea (). However, we were able to detect rifampicin-resistant colonies after scaling up the system to 1L flasks and plating up to 1010 cells following treatment with both 3µg/ml kanamycin and thiourea (data not shown). These results suggest a role for kanamycin-mediated interference with ribosome function and translation, in the absence of oxidative stress, on significantly lowering mutation rate.
To further demonstrate that antibiotic-mediated ROS formation has a mutagenic component, we examined mutation rates under anaerobic growth conditions (see Experimental Procedures for additional details) following treatment of wildtype E. coli with antibiotics or hydrogen peroxide as described above (). We observed mutation rates near untreated levels for all antibiotic treatments tested (). Treatment with 1mM hydrogen peroxide, which results in direct addition of an oxidant, led to an increase in mutation rate relative to the no-drug control under anaerobic growth conditions (), but this increase was considerably smaller than that exhibited under aerobic growth conditions ().
Antibiotic-resistant strains can arise via drug-mediated selection of pre-existing antibiotic-resistant variants that occur naturally within a population (Livermore, 2003
). Antibiotic-induced oxidative stress may be an additional mechanism that allows for the accumulation of mutations that increase resistance to drugs, irrespective of the drug target of the applied antibiotic. To test this, we measured changes in the MIC of wildtype E. coli
over a period of 5 days of selective growth for the following antibiotics: norfloxacin, kanamycin, ampicillin, tetracycline and chloramphenicol. During the growth period, the cultures were exposed to no drug, norfloxacin, ampicillin, or kanamycin (see Experimental Procedures for more details). In all cases, growth in the absence of antibiotics did not change the MIC for any of the drugs tested (data not shown).
Treatment with 25ng/ml norfloxacin led to an increase in the MIC for norfloxacin and kanamycin, respectively (Figure S1A
). The observed increases in MIC following treatment with norfloxacin were concentration dependent (see Supplemental Data
for more details). Treatment of wildtype E. coli
with 3µg/ml kanamycin led to an increase in the MIC for kanamycin and minimal increases in the MIC for norfloxacin and ampicillin, respectively (Figure S1C
). The MIC for tetracycline and chloramphenicol, respectively, did not change (Figure S1C
), indicating that kanamycin treatment may not lead to mutants resistant to other classes of ribosome inhibitors.
Treatment of wildtype E. coli
with 1µg/ml ampicillin for 5 days led to an increase in the MIC to different levels for ampicillin, norfloxacin, kanamycin, tetracycline and chloramphenicol, respectively (). These results show that treatment with a β-lactam can stimulate formation of mutants that are potentially resistant to a wide range of antibiotics. Cultures that had been grown for 5 days in the presence of low levels of ampicillin were shifted to a drug-free environment, and grown without any ampicillin for 2 additional days. The MICs, which were increased after 5 days of ampicillin treatment (), remained elevated and did not change significantly following two days of growth in the absence of ampicillin (Figure S2
). These findings demonstrate that the observed increases in MIC are stable and not due to a transient adaptation to growth in the presence of ampicillin.
Low levels of bactericidal antibiotics can lead to broad-spectrum increases in MIC due to ROS-mediated mutagenesis
To determine if the observed increases in MIC were related to antibiotic-mediated ROS formation, we measured the MIC for ampicillin, norfloxacin, kanamycin, tetracycline and chloramphenicol, respectively, following treatment with no drug or 1µg/ml ampicillin under anaerobic growth conditions. Untreated anaerobic growth had no effect on MIC relative to untreated aerobic growth (data not shown). Following treatment with 1µg/ml ampicillin under anaerobic conditions, we observed almost no increase in MIC for ampicillin, kanamycin, tetracycline or chloramphenicol (). The MIC for norfloxacin exhibited a small increase by day 5 (); however, this change in MIC was much smaller than the increase in MIC for norfloxacin following ampicillin treatment under aerobic growth conditions (). These results suggest that ROS formation due to treatment with low levels of bactericidal antibiotics can lead to mutagenesis and the emergence of bacteria resistant to a wide range of antibiotics.
Drug resistance may not always be uniform throughout a population. Some cells within a population may remain susceptible to the antibiotic whereas other cells display varying degrees of drug resistance (de Lencastre et al., 1993
), a phenomena referred to as heteroresistance. Antibiotic-stimulated, ROS-mediated mutagenesis could be a mechanism that stimulates the formation of a range of mutations that result in varying MICs within a population of cells. We sought to determine if the observed increases in population-level MIC for ampicillin following 5 days of treatment with 1µg/ml ampicillin () exhibited heterogeneity in MIC at the single colony level.
We isolated individual colonies following ampicillin treatment and measured the MIC of each clone to ampicillin. We found that these isolates exhibited a range of resistance to ampicillin (>2.5–12.5µg/ml), with some isolates remaining completely susceptible (≤ 2.5µg/ml) to treatment with this drug (). We also found that the MICs for these isolates to norfloxacin ranged from <100ng/ml (completely susceptible) to ≥ 1000ng/ml (). Although levels of resistance from clinical isolates are typically quite high (with MICs in the range of 10–1000µg/ml for norfloxacin (Becnel Boyd et al., 2009
)), the upper ranges of the MICs for ampicillin or norfloxacin observed here () are near the peak serum concentrations for these drugs (Bryskier, 2005
), indicating that these MICs might be near the limit for the amount of drug a human can tolerate. These data show that heterogeneous increases in MIC to ampicillin arise in E. coli
following treatment with low-levels of ampicillin, and treatment with one drug class can lead to heterogeneous increases in MIC against other classes of antibiotics.
Ampicillin treatment of E. coli results in heterogeneous increases in MIC for ampicillin and norfloxacin
Resistance to multiple antibiotics has been linked to mutations in drug-efflux systems such as the AcrAB multidrug (MDR) efflux pump (George and Levy, 1983
; Ma et al., 1993
), as well as mutations in transcription factors controlling these systems, such as MarA (Alekshun and Levy, 1997
), Rob (Ariza et al., 1995
) and SoxS (Greenberg et al., 1990
). Our results suggest that ROS-mediated DNA damage induced by low levels of bactericidal antibiotics can result in mutations in a wider range of genes, potentially in some unrelated to the applied antibiotic and drug efflux systems. This implies that treatment with ampicillin, for example, may generate mutants that are not ampicillin resistant but are resistant to other antibiotics.
To determine if these types of resistant strains arise, we examined multidrug resistance following 5 days of treatment with 1µg/ml ampicillin or no treatment. Mutants from ampicillin-treated or untreated cultures were selected on plates containing norfloxacin, ampicillin, kanamycin, tetracycline and chloramphenicol, respectively. From this primary selection, we determined cross-resistance to the other four antibiotics via replica plating (see Experimental Procedures for additional details). We found substantially more primary resistant colonies and higher rates of cross-resistance following ampicillin treatment as compared to no treatment (). Ampicillin-selected mutants displayed a range of cross-resistance to the other classes of antibiotics, and showed a strong correlation (89% cross-resistance) with norfloxacin resistance (). We also found that ampicillin-treated cells selected originally on the basis of norfloxacin or kanamycin resistance were only 75% and 63% cross-resistant to ampicillin, respectively (). Interestingly, primary resistance selection with the static drugs tetracycline and chloramphenicol yielded isolates that were always (100%) cross-resistant to ampicillin (); this effect deserves further study. Also of note, ampicillin-treated, kanamycin-resistant strains were found to have very low cross-resistance to tetracycline (7%) and no cross-resistance with chloramphenicol (). This is consistent with previous work showing a lack of cross-resistance to tetracycline or chloramphenicol following selective treatment with aminoglycosides (Grassi, 1979
). While the majority of these multidrug cross-resistant strains exhibit resistance against the treatment drug, ampicillin, our results demonstrate that treatment with ampicillin can also generate mutants that are not resistant to ampicillin, yet are resistant to other classes of antibiotics.
Cross-resistance following ampicillin treatment and primary resistance selection with 5 different classes of antibiotics
We sought to determine if some of the ampicillin-treated, cross-resistant isolates had acquired mutations in specific antibiotic targets or in genes making up the common oxidative damage cell death pathway induced by bactericidal antibiotics (Kohanski et al., 2007
; Kohanski et al., 2008
), or if the observed cross-resistance () was solely a function of altered drug efflux. We examined 6 norfloxacin-resistant isolates, 6 kanamycin-resistant isolates, and the untreated control strain. We sequenced gyrA
, which code for the subunits of DNA Gyrase, the known target of quinolones, rpsL
, which codes for a component of the 30S subunit of the ribosome and has been associated with aminoglycoside resistance, ampC
, which has been associated with ampicillin resistance, icdA
, and iscR
which are genes involved in the common mechanism of cell death, as well as tolC
and its promoter region, and acrA
and its promoter region, which are involved in multidrug efflux.
We found that 3 of the 6 norfloxacin-resistant isolates contained point mutations in gyrA
that resulted in a substitution of glycine for aspartic acid at amino acid 82 in one isolate, and a substitution of tyrosine for aspartic acid at amino acid 87 in two other isolates (). We also found that one of these 6 norfloxacin-resistant isolates, which did not have a mutation in gyrA
, had a point mutation resulting in the conversion of serine to phenylalanine at residue 464 of GyrB (). Interestingly, the point mutations we found in gyrA
are all in the quinolone resistance determining regions of GyrA and GyrB, respectively, and these mutations have been observed in clinical isolates of Bacteroides fragilis
(Oh et al., 2001
), Salmonella enterica
(Weill et al., 2006
) and Pseudomonas aeruginosa
(Mouneimne et al., 1999
Ampicillin treatment leads to the formation of norfloxacin-resistant isolates with mutations in gyrA, gyrB or the acrAB promoter (PacrAB) and kanamycin-resistant isolates with mutations in rpsL or arcA
As noted above, mutations in rpsL have been associated with aminoglycoside resistance. We found that 2 of the 6 kanamycin-resistant isolates had point mutations in rpsL. These mutations led to a frame shift and truncated form of RpsL in both isolates (). It is possible that these mutations contribute to kanamycin resistance in these isolates.
Among the ampicillin-treated, drug-resistant mutants, we did not find any mutations in ampC
(data not shown), a gene associated with ampicillin resistance. We also did not find any mutations in icdA
, or iscR
(data not shown). Interestingly, we did find a single insertion mutation in arcA
in one of the drug-resistant isolates. ArcA is a two-component system transcription factor containing a sensor domain and a DNA binding domain, and the mutation we found results in a truncated ArcA protein that is missing the DNA-binding element of the protein (). We have previously shown that two-component systems are important elements in the common mechanism of cell death, and a knockout of arcA
is more tolerant to treatment with ampicillin and kanamycin compared to norfloxacin (Kohanski et al., 2008
). This isolate is resistant to ampicillin and kanamycin, but not to norfloxacin. This result suggests that mutations leading to low-level antibiotic resistance can occur in genes that are involved in the common mechanism of cell death.
We did not find any mutations in tolC
, the marRA
promoter, or acrA
(data not shown); however, we did find a T-to-A conversion in the promoter upstream of acrA
() in one of the norfloxacin-resistant isolates that also had a mutation in gyrA
(). This promoter mutation occurs within the annotated -35 site of the promoter and the binding site for the repressor transcription factors AcrR and EnvR (Keseler et al., 2005
; Miller et al., 2002
). The observed mutations could reduce the ability of these repressors to bind to the acrAB
promoter, which would result in increased pump expression and drug resistance. These sequencing results demonstrate that ampicillin treatment can lead to the formation of norfloxacin-resistant strains with mutations in DNA Gyrase and/or mutations that can affect drug efflux pump activity, which likely contribute to the emergence of multidrug resistance.
To demonstrate that sub-lethal levels of bactericidal antibiotics can lead to an increase in multidrug cross-resistance in Gram-positive as well as Gram-negative bacteria, we also examined multidrug cross-resistance in Staphylococcus aureus
), following treatment with low-levels of ampicillin (35ng/ml) for 5 days. Previously, we demonstrated antibiotic-mediated ROS formation in S. aureus
(Kohanski et al., 2007
). In the present study, we found substantially more primary resistant S. aureus
colonies and higher rates of cross-resistance following ampicillin treatment as compared to no treatment (). Interestingly, we were unable to enrich for tetracycline- or chloramphenicol-resistant S. aureus
isolates following treatment with low-levels ampicillin as compared with the no-drug treatment. This may be due to the lower level of ROS formation we have observed with S. aureus
(Kohanski et al., 2007
Cross-resistance for S. aureus following ampicillin treatment and primary resistance selection with 5 different classes of antibiotics
To demonstrate that these effects are not limited to lab strains, we considered a clinical isolate of E. coli from a patient with diarrhea (NCDC C771). We examined multidrug cross-resistance in the clinical isolate following treatment with 1µg/ml ampicillin (see Experimental Procedures for more details). As with the wildtype strains, we found substantially more primary resistant colonies and higher rates of cross-resistance in the clinical isolates following ampicillin treatment as compared to no treatment. We also found that ampicillin-treated cells selected originally on the basis of norfloxacin or kanamycin resistance were only 11.5% and 21.5% cross-resistant to ampicillin, respectively (). This further affirms that treatment with ampicillin can generate mutants that are not resistant to ampicillin, yet are resistant to other classes of antibiotics.
Cross-resistance for E. coli clinical isolate NCDC C771 following ampicillin treatent and primary resistance selection with 5 different classes of antibiotics