Intracellular Activity of Antibiotics in a Model of Human THP-1 Macrophages Infected by a Staphylococcus aureus Small-Colony Variant Strain Isolated from a Cystic Fibrosis Patient: Study of Antibiotic Combinations

In a companion paper (H. A. Nguyen et al., Antimicrob. AgentsChemother. 53:1434-1442, 2009), we showed that vancomycin, oxacillin,fusidic acid, clindamycin, linezolid, and daptomycin are poorlyactive against the intracellular form of a thymidine-dependentsmall-colony variant (SCV) strain isolated from a cystic fibrosispatient and that the activity of quinupristin-dalfopristin,moxifloxacin, rifampin, and oritavancin remains limited (2-to 3-log CFU reduction) compared to their extracellular activity.Antibiotic combination is a well-known strategy to improve antibacterialactivity, which was examined here against an intracellular SCVstrain using combinations with either rifampin or oritavancin.Time-kill curve analysis using either concentrations that causeda static effect for each antibiotic individually or concentrationscorresponding to the maximum concentration in human serum showedlargely divergent effects that were favorable when antibioticswere combined with rifampin at low concentrations only and withoritavancin at both low and high concentrations. The natureof the interaction between rifampin, oritavancin, and moxifloxacinwas further examined using the fractional maximal effect method,which allows categorization of the effects of combinations whendose-effect relationships are not linear. Rifampin and oritavancinwere synergistic at all concentration ratios investigated. Oritavancinand moxifloxacin were also synergistic but at high oritavancinconcentrations only. Rifampin and moxifloxacin were additive.This approach may help in better assessing and improving theactivity of antibiotics against intracellular SCV strains.

INTRODUCTION

Top

INTRODUCTION

Small-colony variant (SCV) strains of Staphylococcus aureusare notoriously difficult to eradicate with most commonly usedantistaphylococcal agents (30). Failures favor both selectionand acquisition of antibiotic resistance because SCV strainsare hypermutators (8). In the companion paper (21), we comparedthe intracellular activity of a series of 13 antibiotics againsta methicillin-susceptible thymidine-dependent SCV strain isolatedfrom a cystic fibrosis (CF) patient. We showed that most ofthese antibiotics are poorly active, with only moxifloxacin,rifampin, and oritavancin able to decrease the intracellularinoculum to less than 1% of the initial value.

Antibiotic combination is a well-known strategy both to preventthe emergence of resistant organisms and to increase activityagainst extracellular organisms thanks to additive, and evensometimes synergistic, effects between coadministered drugs(12, 23). Available data suggest that combining antibioticscan improve intracellular activity against both normal and SCVphenotypes of S. aureus strains (2-4). These studies, however,used only a limited number of antibiotics and fixed concentrations,making it difficult to appreciate the exact nature of interactionsbetween combined drugs. We therefore undertook to reexaminethis issue by including a series of commonly used antibiotics.These were combined with either rifampin, systematically usedin most combinations (24), or with oritavancin, a lipoglycopeptidewith marked bactericidal activity toward S. aureus (22, 26);oritavancin is currently in late clinical development (9), andit was one of the most effective drugs in our model (21). Inthe present study, antibiotic combinations were first testedat two fixed concentrations selected to provide meaningful microbiologicaland pharmacodynamic information. The most promising combinationswere then tested using the fractional maximal effect (FME) approach(10, 16), a method that allows analysis of the effects of combinationswhen dose-effect relationships are not linear (21) and thatuses calculated rather than arbitrarily chosen concentrationratios. The study demonstrates that the combination of rifampinand oritavancin allows nearly complete eradication of an intracellularSCV strain, and it calls for a systematic assessment of antibioticcombinations in difficult-to-treat infections caused by thesevariants of S. aureus.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Bacterial strain, intracellular infection, antibiotics, determination of intracellular activity, and statistical analyses. All methods, except those specifically related to the antibioticcombinations described in this paper, were as described in thecompanion paper (21). In brief, experiments were conducted witha stable, thymidine-dependent, fully susceptible SCV strainof methicillin-susceptible S. aureus isolated from a chronicallyinfected CF patient and used to infect THP-1 macrophages. Intracellularinfection was performed as described previously (5), exceptthat the number of CFU was determined after a 48-h incubationof cell lysate on brain heart infusion agar. Intracellular activity(E) was defined as the decrease in the number of CFU recordedat a given time after phagocytosis and addition of antibiotic(s)at specified concentration(s) in comparison with the postphagocytosisinoculum and is expressed in log₁₀ units. A contribution fromextracellular bacteria liberated from lysed macrophages in thesecounts can be ruled out for the following reasons: (i) extracellularbacteria were eliminated by a washing step at the end of theincubation; (ii) cultures were maintained with antibiotics atconcentrations exceeding their MICs, preventing extracellulargrowth; and (iii) no major loss of macrophage viability wasobserved as assessed by the measure of the release of lactatedehydrogenase and by the maintenance or even an increase incell protein (suggestive of cell stasis or growth over the incubationperiod).

These E values are negative because they pertain to decreasesin bacterial counts; a greater activity is therefore associatedwith a more negative E value. Since this is rather counterintuitive,we will use the absolute values of E when comparing activitiesthroughout this paper. (Note however that the tables in boththis paper and the companion paper show actual values and notabsolute values.)

Assessment of the effect of antibiotic combinations. Two successive approaches were followed. In the first instance,antibiotics were combined as follows: (i) at their respectivestatic concentrations (i.e., the extracellular concentrationcausing no apparent change in the intracellular CFU count comparedto the postphagocytosis inoculum [C_s], as obtained from previousdose dependence experiments with each antibiotic [see Table1 in this paper for values and Table 2 in reference 21 for pharmacologicaldescriptions of concentration-effect relationships]) and (ii)at an extracellular concentration corresponding to the maximalserum concentration (total drug) observed in patients afteradministration of conventional doses of the corresponding antibioticto humans ([C_max]see Table 2 in this paper for individual valuesand Table 1 in reference 21 for references). To check that theincreases in activity seen when antibiotics are combined attheir C_max values were not due to a carryover effect upon platingof cell lysates containing high intracellular concentrationsof antibiotics, we compared bacterial counts from cultures exposedto cell lysates incubated for 24 h with antibiotics at theirC_max values; samples were then either left untreated or treatedwith an equal volume of 25 mg/ml activated charcoal suspension(6). No difference was seen between samples treated with charcoalor left untreated or between samples exposed to cell lysateschallenged by antibiotics or to unchallenged lysates, rulingout that any carryover effect could take place under our experimentalconditions.

View this table:
[in this window]
[in a new window]

TABLE 1. Comparative intracellular activities of antibiotics alone or combined with rifampin or with oritavancin against an SCV strain exposed for the indicated times at a fixed extracellular concentration giving rise to a static intracellular effect when tested alone^a

View this table:
[in this window]
[in a new window]

TABLE 2. Comparative intracellular activities of antibiotics alone or combined with rifampin or with oritavancin against an SCV strain exposed for the indicated times at a fixed extracellular concentration corresponding to the human C_max^a

In a second instance, selected combinations were examined indetail using the FME approach (10, 16). In the present study,all measurements were made after a 24-h incubation, and thedecrease in the intracellular bacterial counts (E) was usedas the end point. The E_max, i.e., the maximum decrease in theinoculum, was extrapolated from Hill's function for an infinitelylarge antibiotic concentration after a 24-h incubation (in logdecrease CFU at 24 h; values are 1.72 for rifampin, 1.32 formoxifloxacin, and 0.43 and 3.13 for oritavancin, where the twosuccessive values reflect a bimodal effect [see Table 2 in reference21]); the E_max was taken as the activity level correspondingto an FME of 1 (all E_max values used in the present study correspondto bacterial counts that are above the limit of detection).The concentrations of each antibiotic expected to yield an FMEof 0.1, 0.3, 0.5, 0.7, or 0.9 (C_xp) when given alone were thencalculated using equation 1:

Formula 1 (1)
where EC₅₀ is the extracellular antibiotic concentration (mg/liter)causing a reduction of the inoculum to halfway between the numberof CFU extrapolated for an infinitely low antibiotic concentration(E_min) and E_max in 24 h (EC₅₀ values, in mg/liter, are 0.02for rifampin, 0.3 for moxifloxacin, and 0.6 and 19 for the twosuccessive parts of the regressive function fitted to the oritavancindata [converted in mg/liter from EC₅₀ values expressed in multiplesof the MIC as shown in Table 1 of the companion paper]). Thevalidity of the model was tested by running experiments at eachof these concentrations and comparing the observed values ofFME (FME_obs; defined as E/E_max) to the expected FME (FME_xp).To assess the effect of combining two antibiotics, antibioticA was added at five increasing concentrations (C_xpA) expectedto yield FMEs of 0.1, 0.3, 0.5, 0.7 and 0.9, whereas antibioticB was added at concentrations (C_xpB) expected to provide thecomplement (0.9, 0.7, 0.5, 0.3, and 0.1), so that the expectedtotal (FME_xp(A_+B₎) should always be equal to 1. The level ofintracellular activity expected for an additive effect (E_xpadditive) of antibiotics A and B was then calculated using Katzper'sformula (10), shown as equation 2:

Formula 2 (2)
and compared to the measured level of activity(E) to calculate the actual FME (FME_obs(A_+B₎) of the combinationusing equation 3:

Formula 3 (3)

Values of FME_obs(A_+B₎ were then plotted as a function of theconcentration ratios of the two drugs expected to give an FMEof 1 (from 0.1/0.9 to 0.9/0.1), together with the values ofthe FME_obs of each antibiotic alone using its actual concentration.In these graphs, a synergistic effect will yield an ordinatevalue of >1, an additive effect yields a value of 1, an indifferenteffect yields a value <1 but higher than the values of theof FME_obs of the corresponding antibiotics alone, and an antagonisticeffect has a value of <1 and below the FME_obs of the correspondingantibiotic alone (10).

RESULTS

Top

RESULTS

Combinations at fixed concentrations. In a first approach, antibiotics were combined at two fixed,arbitrarily chosen concentrations, namely, their static concentrations(C_s) or at a concentration corresponding to their respectiveC_max (total drug). The results are first illustrated in Fig.1 for studies examining the changes in the intracellular bacterialcounts over time for rifampin and oritavancin taken as typicalexamples of a combination yielding a beneficial effect comparedto what was obtained with each of these antibiotics alone. Basedon the experiments made with the static concentrations of eachdrug (Fig. 1), the benefit of the combination clearly appearsas a decrease in the number of CFU of about 1.2 logs at 5 or24 h compared to no change for each antibiotic used alone (themodel did not allow us to examine longer exposure times becauseof an overwhelming growth of extracellular bacteria after 24h as a result of the low antibiotic concentrations used). Thebenefit of the combination was also evident when the two antibioticswere used at extracellular concentrations corresponding to theirrespective C_max values (Fig. 1). Of note is that the combinationof rifampin and oritavancin decreased the 72-h CFU count tothe limit of detection (4.5 logs), an effect that, so far, hasnever been seen in our model for an antibiotic given alone,regardless of its concentration and the strain examined (5,21).

View larger version (15K):
[in this window]
[in a new window]

FIG. 1. Influence of time on the rate and extent of activity of rifampin (RIF), oritavancin (ORI) and of their combination (RIF+ORI) against intracellular SCVs upon incubation at a fixed extracellular concentration corresponding to the following: (i) a static effect (C_static as determined in the companion paper (21) (values are 0.003 mg/liter for rifampin and 0.14 mg/liter for oritavancin) or (ii) their maximal concentration (total drug) observed in humans after administration of conventional doses (C_max) (values are 18 mg/liter for rifampin [1] and 25 mg/liter for oritavancin [29]) when drugs are tested alone. The ordinate is the change in the number of CFU per mg of cell protein (log scale). Values are means ± standard deviations (n = 3); when not visible, the error bars are smaller than the symbols.

Table 1 summarizes the results obtained with all antibioticsexamined in the present study, comparing their activities whenthey are used alone or in combination with rifampin or oritavancinwith their activities when they are used at their static concentrations(the corresponding kill curves are shown in the left panelsof Fig. SP1 in the supplemental material). The addition of rifampinwas clearly beneficial for every antibiotic tested, with thegreatest reductions in bacterial counts observed with oxacillin,moxifloxacin, gentamicin, and oritavancin (ranked from smallerto greater effects) at 24 h (and also for fusidic acid but onlyat 5 h). The addition of oritavancin was also beneficial, butthe gain in activity was lower at 24 h than observed in combinationsinvolving rifampin, with differences between antibiotics thatwere either modest or statistically nonsignificant, except forthe combination oritavancin-rifampin.

Table 2 shows the results obtained for antibiotics combinedat concentrations mimicking the human C_max (total drug) andfor postphagocytosis incubation times of 24 and 72 h (the correspondingkill curves are shown in the right panels of Fig. SP1 in thesupplemental material). All combinations with rifampin led toa degree of intracellular killing similar to that obtained withrifampin at 24 h (1- and 1.5-log CFU decrease from the postphagocytosisinoculum) except for quinupristin-dalfopristin and, to a largerextent, for oritavancin, with which a greater degree of intracellularactivity was obtained. Similar observations were made at 5 h,in the sense that the reduction in bacterial counts was of thesame order of magnitude (about 1 log) for rifampin or oritavancinalone or in combination with the other antibiotics (see FigureSP1 in supplemental material). At 72 h, slightly less activitycompared to rifampin alone was observed for most drugs, butthis difference reached 1 log or more for fusidic acid, clindamycin,and linezolid. Notably, the combination with oritavancin wasthe only one to show higher activity than rifampin alone, reachingthe limit of detection (–4.5 logs). When we then examinedthe combinations with oritavancin, we observed a higher activityfor rifampin, gentamicin, moxifloxacin, and quinupristin-dalfopristin,with the increase in activity compared to oritavancin alonereaching for these four drugs $~$ 0.2 log at 5 h (see Figure SP1in the supplemental material), $~$ 0.5 log at 24 h, and at least1 log at 72 h, respectively.

The next series of experiments aimed at further characterizingthe nature of the interaction within combinations. To this effect,we used the FME method (10, 16), focusing on the combinationsbetween rifampin, oritavancin, and moxifloxacin since thesedrugs proved very active in our model when used alone, withclear indications of a favorable effect of their combinationswhen tested at fixed concentrations (see above). Maximal effectsused for calculating concentrations were a 1.72-log and 1.32-logCFU decrease for rifampin and moxifloxacin, respectively (21).As studies with oritavancin alone revealed a bimodal dose-effectrelationship (21), we considered for this case two E_max values(at 0.43- and 3.13-log reduction), corresponding to the twosuccessive plateaus observed in these experiments. We firstvalidated the approach by examining to what extent the observedvalues of FME (FME_obs) were linearly correlated with the concentrationof these antibiotics when tested alone. This is shown in Fig.2 with an overall correlation coefficient (R²) of 0.80 for allfour conditions (oritavancin using E_max values correspondingto low and high concentrations; moxifloxacin; and rifampin)and a slope of 1.10 ± 0.13 between the observed and expectedFME values. The antibiotics were then combined two by two, andthe resulting FME values of the combinations are shown in Fig.3. The combination of rifampin with oritavancin was synergisticat all concentrations tested when examining the range of oritavancinconcentrations corresponding to the first plateau of activitywhen tested alone (up to about 5 mg/liter). This combinationmaintained a synergistic effect at higher oritavancin concentrations(in the range of 8 to 45 mg/liter, with rifampin varying from0.04 to 0.007 mg/liter) but became additive above and belowthese values. The combination of moxifloxacin with oritavancinwas indifferent for the lowest oritavancin concentrations investigated(FME of <1 but above the curve of the drugs alone) but becameadditive (FME close to 1) at higher concentrations, with a synergisticeffect (FME of >1) when moxifloxacin and oritavancin wereused at a 0.3/19 mg/liter concentration ratio. Combining moxifloxacinwith rifampin showed essentially an additive effect (FME wasclose to 1 over the whole range of concentrations ratios investigated).

View larger version (21K):
[in this window]
[in a new window]

FIG. 2. Correlation between observed and expected FMEs of four antibiotics toward intracellular SCVs when used alone. Antibiotics were added at concentrations calculated to yield FME values of 0.1, 0.3, 0.5, 0.7, and 0.9, and the actual FME calculated based on the observed activities at 24 h was compared to the maximal activities as determined in dose-dependence experiments [E_max values of –1.72 log CFU for rifampin and –1.32 log CFU for moxifloxacin; for oritavancin, two successive E_max values of –0.43 log CFU at low (L) and –3.13 log CFU at high (H) concentrations were considered based on the observation of a bimodal concentration-effect relationship (see Table 2 in reference 21)].

View larger version (32K):
[in this window]
[in a new window]

FIG. 3. FME plots of antibiotics against intracellular SCV. In each graph, the abscissa shows the concentrations of antibiotics tested, indicated above the graphs, either alone or in combination and expressed as follows: (i) the five expected FME values examined for antibiotics alone (0.1 to 0.9 or 0.9 to 0.1) and the ratios of the expected FME values contributed by each antibiotic (from 0.1/0.9 to 0.9/0.1) for antibiotic combinations; and (ii) the corresponding concentrations (in mg/liter) used. The ordinate shows the value of the observed FME for each antibiotic alone (open symbols) or for the combination [FME_obs(A_+B₎; shown as closed symbols in A and B and as stars in C]. Values of FME_obs(A_+B₎ that are >1 denote a synergistic effect; values equal to 1 indicate an additive effect; values of <1 but higher than FME of the antibiotics alone (FME_obsA or FME_obsB) indicate an indifferent effect (a value of <1 and below FME_obsA or FME_obsB would indicate an antagonistic effect, which was not observed here) (10). For oritavancin, two ranges of concentrations were examined based on the observation of a bimodal concentration-effect relationship with two successive E_max values of –0.43 log CFU at low (L) and –3.13 log CFU at high (H) concentrations (see Table 2 in reference 21).

DISCUSSION

Top

DISCUSSION

The present study contributes to ongoing systematic effortsat defining conditions that are likely to improve the therapeuticoutcome of intracellular infections due to SCV strains by meansof antibiotic combinations (2, 3). The model used in this workis derived directly from the model that we used previously toestablish key pharmacological properties of antistaphylococcalantibiotics against intracellular S. aureus with normal-phenotypeand SCV strains, using a well-controlled environment and minimizinginterferences by host-related factors (5, 21).

The first key observation made here is that the combinationof either rifampin or oritavancin with a series of commonlyused antistaphylococcal agents may produce largely divergenteffects, based on the following observations: (i) globally favorableeffects were observed with rifampin at low concentrations only(and its activity was partly diminished when combined with thesame antibiotics at high concentrations); and (ii) oritavancinproduced favorable effects at both low and high concentrations.A survey of the adjunctive use of rifampin for the treatmentof S. aureus infections suggests antagonism or indifferencewith β-lactams, fluoroquinolones, linezolid, or daptomycin;positive or negative effects, depending on the model, for vancomycin,fusidic acid, or clindamycin; and synergy only for quinupristin-dalfopristin(24). These conclusions are largely supported in our model aswell, except that we did not observe a marked advantage whenwe combined rifampin with quinupristin-dalfopristin. Our resultsalso show that the concentrations at which these effects aremeasured were critical and that favorable effects with rifampinmay only clearly appear when the drug is used at suboptimalconcentrations. Our observation that the combination of oritavancinwith other antistaphylococcal drugs is systematically favorableis in line with observations made with S. aureus strains withvarious resistance phenotypes when they are exposed to combinationsof oritavancin with gentamicin, vancomycin, or rifampin (6,7, 18).

A second key observation is that the combination of rifampinwith oritavancin was markedly and almost always synergistic,with a global effect suggesting that near eradication of anintracellular SCV strain could be obtained. If confirmed inother models, this observation could have considerable valuefor the following reasons: (i) persistence of SCVs is consideredas one of the significant causes of persistence and relapsesin staphylococcal infections (31); (ii) these variants couldoften be intracellular in vivo (30) and appear only poorly susceptibleto antibiotics when used alone in our model (21). One caveatof our observations, however, is that eradication might requireextracellular concentrations that may not be attainable in vivounder clinically acceptable conditions. It is nevertheless pertinentthat a large decrease in the intracellular inoculum was observedat lower, more physiologically relevant concentrations of rifampinand oritavancin.

The model used in the present study, however, has some featuresthat may limit its significance. As analyzed in our previouspublications (5, 14, 15), these limitations include a lack ofdynamic aspect (i.e., the antibiotics are maintained at fixedconcentrations throughout the observation period), which couldbe important here in view of the duration of the experiments,as well as the impossibility of assessing the importance ofprotein binding, a point that will need careful attention fororitavancin since this antibiotic is highly protein bound (25,32). Our study is also limited to a single strain, preventingus from drawing general conclusions on the extent of synergyreached, which may depend on the strain, its susceptibilityto antibiotics, its intracellular growth, or the metabolic defectresponsible for its SCV character, as illustrated by comparingthe data generated here with those obtained for two hemin-dependentSCV strains with higher levels of intracellular growth (2).

Findings from our study provide further impetus to investigatethe mechanisms that result in synergistic activity against intracellularbacteria. We may, however, speculate that the systematicallyfavorable effects seen with oritavancin could be related (i)to its intense bactericidal effect, probably due to membrane-destabilizingproperties (11, 17) that may favor access to target for otherantibiotics (6), and/or (ii) to the high concentration thatoritavancin reaches in the phagolysosomes (28) where SCV strainsalso locate (21). The negative effect exerted by all antistaphylococcaldrugs, except oritavancin, toward rifampin when tested at highconcentrations also requires further explanation. Unfortunately,we could not obtain direct information about drug stabilityand the absence of drug-drug interactions within macrophagesduring the prolonged exposure times needed to assess antibioticactivity against intracellular SCVs. However, the fact thatrifampin is stable for several days at 37°C in a susceptibilitytesting medium (13), the lack of specificity with respect tothe pharmacochemical class of the drugs tested, and the enhancingeffect of oritavancin on the activity of several of these antibioticsmake interpretations based on chemical alteration or physicalinactivation of the drugs unlikely.

In spite of these uncertainties, and pending additional datagathered from more pertinent models, our data highlight theinterest of combining antibiotics having an appropriate pharmacokinetic/pharmacodynamicprofile against S. aureus (5, 27) and showing activity againstnongrowing bacteria (19, 20) to successfully control and reduceintracellular infection by SCVs. This study may trigger furtherevaluation of antibiotic combinations against intracellularbacteria and ultimately may help in the design of pertinentclinical trials.

ACKNOWLEDGMENTS

This work was supported by the Institut d'encouragement de laRecherche Scientifique et de l'Innovation de Bruxelles (IRSIB)/Instituutter Bevordering van het Wetenschappelijk Onderzoek en de Innovatievan Brussel (IWOIB) of the Brussels Capital Region, within theframework of the Research in Brussels program, by the Fondsde la Recherche Scientifique Médicale (grants 3.4.597.06and 3.4.639.09), and by a grant-in-aid from Targanta TherapeuticsCorp. H.A.N. was successively a postdoctoral fellow of the IRSIB/IWOIBand of the Fonds de la Recherche Scientifique-Fonds Nationalde la Recherche Scientifique (FRS-FNRS); F.V.B. is Maîtrede Recherches of the Belgian FRS-FNRS.

We are grateful to J. J. Schentag (University of Buffalo, Buffalo,NY) for valuable advice in the design of our experiments anduseful discussions. M. C. Cambier and C. Misson provided dedicatedtechnical assistance.

FOOTNOTES

* Corresponding author. Mailing address: Unité de Pharmacologie Cellulaire et Moléculaire, Université Catholique de Louvain, UCL73.70 avenue Mounier 73, 1200 Brussels, Belgium. Phone: 32 2 764 73 78. Fax: 32 2 764 73 73. E-mail: francoise.vanbambeke{at}uclouvain.be

Published ahead of print on 2 February 2009.

Supplemental material for this article may be found at http://aac.asm.org/.

REFERENCES

Top