Antimicrobial synergy testing is used to evaluate the effect of two or more antibiotics used in combination and is typically performed by one of two methods: the checkerboard array or the time-kill assay. Here, we present an automated, inkjet printer-assisted checkerboard array synergy technique and a classic time-kill synergy study.
As rates of multidrug-resistant (MDR) pathogens continue to rise, outpacing the development of new antimicrobials, novel approaches to treatment of MDR bacteria are increasingly becoming a necessity. One such approach is combination therapy, in which two or more antibiotics are used together to treat an infection against which one or both of the drugs may be ineffective alone. When two drugs, in combination, exert a greater than additive effect, they are considered synergistic. In vitro investigation of synergistic activity is an important first step in evaluating the possible efficacy of drug combinations. Two main in vitro synergy testing methods have been developed: the checkerboard array and the time-kill study. In this paper, we present an automated checkerboard array method that makes use of inkjet printing technology to increase the efficiency and accuracy of this technique, as well as a standard manual time-kill synergy method. The automated checkerboard array can serve as a high-throughput screening assay, while the manual time-kill study provides additional, complementary data on synergistic activity and killing.
The checkerboard array is a modification of standard minimum inhibitory concentration (MIC) testing, in which bacteria are incubated with antibiotics at different concentration combinations and evaluated for growth inhibition after overnight incubation. Manual performance of the checkerboard array requires a laborious and error-prone series of calculations and dilutions. In the automated method presented here, the calculation and dispensing of required antibiotic stock solution volumes are automated through the use of inkjet printer technology. In the time-kill synergy assay, bacteria are incubated with the antibiotics of interest, both together and individually, and sampled at intervals over the course of 24 h for quantitative culture. The results can determine whether a combination is synergistic and whether it is bactericidal, and provide data on inhibition and killing of bacteria over time.
The spread of multidrug-resistant (MDR) bacterial pathogens, particularly MDR Gram-negative bacteria such as carbapenem-resistant Enterobacteriaceae (CRE), has left clinicians with increasingly limited options for successful anti-infective therapy1, a problem exacerbated by the sluggish pace of novel antibacterial drug discovery2,3. Antimicrobial synergy, in which two drugs used in combination exert a greater-than-additive effect, offers the possibility of salvaging existing antibiotics for use in treatment of MDR bacteria, even when these bacteria are resistant to one or both of the antibiotics individually. The techniques described in this paper provide two complementary methods of in vitro synergy testing that, when used together, allow investigators to efficiently screen antimicrobial combinations of interest for evidence of synergistic activity (the automated checkerboard array method) and then to further evaluate the kinetics of inhibition and killing demonstrated by promising combinations identified in the screening stage (the manual time-kill method).
One of the most commonly used methods of in vitro synergy testing is the checkerboard array assay, a modification of minimum inhibitory concentration (MIC) testing in which the inhibitory activity of two different antibiotics against a bacterial isolate are tested over a range of concentration combinations4,5. If the two drugs exert greater than additive activity when used together, the combination is considered synergistic6. However, setting up a checkerboard array manually involves a series of calculations and diluting and pipetting steps that are laborious and vulnerable to human error. These constraints have had the effect of limiting the use of synergy testing primarily to the retrospective evaluation of small numbers of antibiotic combinations and bacterial isolates, and results have not always been consistent among studies7,8,9,10,11. Furthermore, the complexity of synergy testing has contributed to its unavailability in the clinical microbiology laboratory and to the virtual absence of in vitro synergy testing data from clinical studies of combination therapy12,13.
In order to increase the efficiency and throughput of the checkerboard array method, we made use of an automated MIC testing technique previously developed in our laboratory that uses inkjet printing technology to precisely and consistently dispense small volumes of antibiotic stock solution into wells in a microtiter plate14. The platform obviates the need for complex calculations and multiple pipetting steps. The associated software calculates and dispenses appropriate volumes of antibiotics to create a two-dimensional checkerboard array if the user simply inputs the desired concentration range and stock solution concentration of the antibiotics. We initially tested this method against a collection of CRE isolates15 and subsequently have focused on testing colistin-containing combinations for activity against colistin-resistant isolates16. Colistin is a drug of last resort generally reserved for use in the treatment of MDR Gram-negative pathogens17,18, and colistin resistance renders already MDR bacteria nearly pan-resistant19, making them ideal candidates for the development of novel therapeutic strategies using drugs to which they are insensitive individually. We found that the combination of colistin and the protein synthesis inhibitor antibiotic minocycline had a very high rate of synergy, even against strains that were resistant to each of these drugs individually, presumably because colistin exerts a subinhibitory permeabilizing effect on even colistin-resistant bacteria. We have chosen this combination to use as an example in this paper. Of note, synergy testing can also be used to evaluate for enhanced efficacy of two drugs which are both effective individually.
The automated checkerboard array method facilitates rapid, high-throughput synergy testing. However, the checkerboard array method does have limitations. As a modified MIC assay, it provides data only on inhibition of bacterial growth and not on killing, and it does not provide data on antibiotic effects over time. By contrast, manual performance of time-kill synergy assays is more labor intensive but provides information on both inhibition and killing over a 24 h time course20,21. We used time-kill analysis on a smaller number of isolates to confirm our checkerboard array results and to determine whether the synergistic combinations we identified were also bactericidal.
Both checkerboard array and time-kill synergy methods provide valuable information on the activity of drug combinations, and are particularly useful in evaluating potential novel therapeutic options for highly resistant bacterial pathogens. The methods also have inherent limitations. The standard microbroth dilution MIC method has a known expected error range of 1 two-fold dilution22, which is increased when two drugs are tested together in a checkerboard array. The standard definition of synergy, which considers a combination synergistic only if the drugs are active together at one-fourth their respective MICs6, takes into account this expected variability, but such variability (which is thought to result from a combination of biological and technical fluctuations23) inevitably generates uncertainly about the reliability of synergy results. The lack of established quality-control standards for synergy testing is also a current limitation. Perhaps the most significant limitation of all synergy testing methods is the lack of established correlations between in vitro results and clinical outcomes when combinations are used to treat patients24. Simpler and more rapid synergy testing methods, such as the automated checkerboard array method described here, may facilitate the integration of in vitro synergy testing within clinical trials or other evaluations of patient outcomes in order to better characterize the relationship between in vitro and in vivo effects in the future.
The automated checkerboard array method that we present here offers an option for high-throughput screening of a variety of combinations and allows for quick evaluation of unusual, "high risk-high reward" combinations without a major investment of time and resources. The time-kill method, which we subsequently demonstrate, can provide additional supportive information on the synergistic activity of the combination and can help to characterize its bactericidal activity and antibacterial kinetics.
CAUTION: Use appropriate safety procedures when working with bacteria. Wear gloves and a lab coat at all times. Perform work in a biosafety cabinet if aerosols will be generated or working with high risk pathogens.
NOTE: Twenty to 24 h before starting experiments, streak out the bacterial isolate(s) to be tested (from a colony-purified, minimally passaged stock frozen at -80 °C in tryptic soy broth with 50% glycerol stock) onto a blood agar plate. Incubate the plate at 35 °C in ambient air.
1. Inkjet Printer-assisted Automated Checkerboard Array Synergy
2. Time-kill Synergy Testing
Figure 1A presents a grid from a checkerboard array synergy experiment in which minocycline in concentrations of 0-32 μg/mL was combined with colistin at concentrations of 0-16 μg/mL and tested against E. coli strain FDA-CDC 0494. The values represent spectrophotometric readings at optical density 600 nm (OD600). Wells with OD600 values below 0.07 (which corresponds to no growth by visual inspection) are shaded red, while wells with OD600 values 0.07 (which corresponds to growth by visual inspection) are shaded green. For each drug, the minimum inhibitory concentration (MIC; bolded) is the lowest concentration of drug that inhibits bacterial growth. For minocycline, this is 32 μg/mL, and for colistin, it is 8 μg/mL. The shading is retained in Figure 1B, but values within the wells in which growth is inhibited are replaced by fractional inhibitory concentration index (FICI) values. These are determined as follows: in each well, the fractional inhibitory concentration index (FIC) of each drug is calculated by dividing the concentration of antibiotic in that well by the drug's MIC, and the FICI is calculated by summing the two FICs. Wells with an FICI value of 0.5, which is considered the cutoff for synergy, are indicated with a broken-line border, and the well with the lowest FICI value (0.094) is bolded. Because the minimum FICI value is in the synergistic range, the combination is considered synergistic.
Figure 2A and Figure 2B show grids analogous to those in Figure 1A and Figure 1B, but in this case the combination does not demonstrate synergy against the isolate tested (K. pneumoniae isolate BIDMC 4), because the minimum FICI at which growth is inhibited is 1, which is >0.5.
Figure 3 illustrates the optical density readings from a checkerboard synergy grid in which several skipped wells occurred (Enterobacter cloacae complex isolate BIDMC 27). Skipped wells are wells in which bacterial growth is inhibited despite the presence of bacterial growth in adjacent wells with higher concentrations of antibiotic. This phenomenon, which is known to occur in standard MIC testing as well, is likely due to biological variability in bacterial growth characteristics from well to well and to the sensitivity of some antibiotics to small differences in bacterial inoculum23,31,32. If more than one skipped well occurred in a checkerboard array, we discarded the results and repeated the assay.
Figure 4 presents examples of time-kill synergy results of three combinations tested against K. pneumoniae isolate BIDMC 32. Colony counts are indicated in a logarithmic scale on the y-axis and time, in hours, on the x-axis. The difference between the starting inoculum in the tube containing the drug combination and the concentration of bacteria in that tube at 24 h is illustrated by the red bar and number, while the difference between the concentration of bacteria at 24 h between the tube containing the combination and the tube containing the most active single agent alone is illustrated by the blue bar and number. Figure 4A shows results from the combination of colistin and minocycline; this combination was synergistic (difference between concentrations of bacteria exposed to combination and to most active agent alone ≥2 log10 CFU/mL at 24 h) and bactericidal (decline from starting inoculum to concentration at 24 h ≥3 log10 CFU/mL). Figure 4B shows results from the combination of colistin and clindamycin, a combination that was synergistic but was not bactericidal. This combination inhibited growth of the bacteria, which neither drug did alone, but did not kill them. Figure 4C shows results from the combination of colistin and erythromycin, which was neither bactericidal nor synergistic.
Figure 1: Checkerboard array results demonstrating synergy (minocycline + colistin tested against E. coli strain FDA-CDC 0494). (A) Spectrophotometric readout and growth interpretation of a checkerboard array. Values in cells are optical density readings at 600 nm (OD600). Cells with OD600 values below 0.07 (corresponding to no growth by visual inspection) are shaded red, while cells with OD600 values 0.07 (corresponding to growth by visual inspection) are shaded green. (B) Fractional inhibitory concentration index (FICI) calculation. Shading indicating growth or no growth has been retained. Values for colistin and minocycline along x– and y-axes, respectively, now represent the fractional inhibitory concentration (FIC), or the ratio of the concentration of the drug in that column or row to the minimum inhibitory concentration (MIC) of that drug alone. The value in each cell is the FICI, or the sum of the FICs of the two drugs in that well. The large broken line-bordered box encloses wells with an FICI of 0.5. The thick-bordered cell indicates the well with the lowest FICI in which growth is inhibited, or the minimum FICI. Because the minimum FICI is 0.5, the combination is considered synergistic. Please click here to view a larger version of this figure.
Figure 2: Checkerboard array results of a combination that does not demonstrate synergy (minocycline + colistin tested against K. pneumoniae isolate BIDMC 4). (A) Optical density values at 600 nm and growth interpretation of checkerboard array results as described for Figure 1A. (B) Fractional inhibitory concentration index (FICI) calculation as described for Figure 1A. Because the minimum FICI is >0.5, the combination is not considered synergistic. Please click here to view a larger version of this figure.
Figure 3: Checkerboard array results that are uninterpretable due to skipped wells (minocycline + colistin tested against Enterobacter cloacae complex isolate BIDMC 27). Optical density values at 600 nm and growth interpretation of checkerboard array results as described for Figure 1A. Several skipped wells, in which bacterial growth is inhibited despite the presence of growth in adjacent wells with higher concentrations of antibiotic, are demonstrated. Results are not interpretable, and experiment needs to be repeated. Please click here to view a larger version of this figure.
Figure 4: Time-kill synergy results of three combinations tested against K. pneumoniae isolate BIDMC 32. Colony counts are indicated in a logarithmic scale on the y-axis and time, in hours, on the x-axis. The difference between the concentration of bacteria in the combination at 24 h and the starting inoculum in the tube is illustrated by the red bar and number. If the decline from starting inoculum to concentration at 24 h is ≥3 log10 CFU/mL, the combination is considered bactericidal. The difference between the concentration of bacteria at 24 h between the tube containing the combination and the tube containing the most active single agent alone is illustrated by the blue bar and number; if there is ≥2 log10 CFU/mL reduction, the combination is considered synergistic. (A) Colistin (CST) + minocycline (MIN), a combination that is both synergistic and bactericidal. (B) Colistin + clindamycin (CLI), a combination that is synergistic but not bactericidal. (C) Colistin + erythromycin (ERY), a combination that is neither synergistic nor bactericidal. These results were initially published as part of a study of the synergistic activity of colistin-containing combinations against colistin-resistant Enterobacteriaceae, in which we demonstrated that colistin was synergistic with a number of antibiotics that are active individually only (e.g. clindamycin) or primarily (e.g. erythromycin) against Gram-positive bacteria16. (Note that erythromycin was synergistic by checkerboard array against the strain shown, but not by time-kill, so it has been selected here as an example of a non-synergistic combination.) We hypothesized that colistin, which is known to act by permeabilization of the Gram-negative outer membrane, exerts a sub-inhibitory permeabilizing effect on colistin-resistant Gram-negative bacteria, allowing entry of drugs such as clindamycin that normally cannot enter the Gram-negative cell. Panel (A) of this figure has been modified from Brennan-Krohn, Pironti, and Kirby 201816, copyright © American Society for Microbiology, Antimicrobial Agents and Chemotherapy, 62(10), 2018, pii: e00873-18, doi: 10.1128/AAC.00873-18. Please click here to view a larger version of this figure.
The two methods described here both provide information about the activity of antimicrobials used in combination compared to their individual activity. The automated, inkjet printer-assisted digital dispensing method is an adaption of the method described in the Clinical Microbiology Procedures Handbook33, while the time-kill method more closely follows the corresponding protocol from the same reference34.
In the checkerboard array method, calculations to determine the necessary volume of antimicrobial stock to add to each well as well as the dispensing of these volumes is automated, thus eliminating some of the major potential sources of error encountered in a manual checkerboard array. It is still essential, however, that the investigator determines that original stocks are made at the intended concentration and that goal final concentrations are entered into the D300 software correctly. Adding the antimicrobial suspension to wells in a 384-well plate can be challenging at first and requires care to ensure that pipette tips enter the appropriate wells and that liquid does not splash up the edges of the wells. An automated liquid handler can be used in place of a hand-held multichannel pipette to increase the speed and accuracy with which the bacterial suspension is added to wells. As described in the protocol, the D300 requires the addition of the surfactant, polysorbate 20 (P-20), for proper liquid handling. A different surfactant, polysorbate 80, at a concentration of 0.002%, has been noted to lower colistin MICs for organisms with colistin MICs of <2 µg/mL in standard broth microdilution assays.35,36 Our laboratory previously demonstrated that P-20 at concentrations up to 0.0015% had no effect on D300-assisted MIC results in comparison with reference BMD14. In the assay example presented here, the maximum P-20 is concentration is 0.0014%.
One problem we encountered with some checkerboard array assays was a large number of skipped wells. This occurred at a disproportionate rate with certain antibiotics. Specifically, in a screen of combinations against a collection of carbapenem-resistant Enterobacteriaceae, we found that while 49 of 521 trials (9.4%) were unusable due to multiple skipped wells, 2 of the 12 antibiotics tested (fosfomycin and cefepime) accounted for 46 of these trials (94%). Such increased rates may be more likely in drugs that are particularly susceptible to the inoculum effect31,32,37. Of note, CLSI does not recommend testing fosfomycin in broth dilution25 due to concerns about the reliability of results with this method, which may explain the unreliable results seen with this drug. Some modifications can be made to automated checkerboard method according to investigator preference. Antimicrobials can be dispensed into plates already containing bacterial suspension, rather than into empty wells, if this is preferable for reasons of workflow within the laboratory. While 384-well plates were used here, the method can also be carried out in 96-well plate assays with appropriate modification of well volume. The use of a 96-well plate format may help in reducing skipped wells for antibiotics that are particularly sensitive to small changes in inoculum. When calculating FICI, there may be situations where the MIC is off-scale (i.e., higher than tested), including situations where the drug being tested has no activity individually against the type of organism being tested. In these cases, the FIC can be calculated based on assuming the MIC is one dilution higher than the highest concentration tested. This is the most conservative strategy, as it assumes the maximal possible FIC value for any dilution where inhibition is observed during synergy testing. For example, if the actual MIC were instead two doubling dilutions above the highest concentration tested, then the corresponding FICs would be two-fold lower than the conservative assignments, and so on.
In order to accurately assess the bactericidal activity of drugs in a time-kill assay, it is essential that cultures be in logarithmic-phase growth, particularly when cell-wall active antibiotics are being tested28. For the rapidly-growing bacteria used in this example (K. pneumoniae), 3 hours of incubation with shaking was appropriate to reach this growth phase, but different amounts of time may be necessary for different organisms. In general, the culture should appear visibly but not heavily turbid. The appropriate amount of time can be determined by constructing a growth curve with colony counts taken at serial time points (e.g., every 30 min for 4-6 h)38. The intended starting inoculum in the time-kill study is also important. The target concentration of the starting inoculum is approximately 5 x 105 to 1 x 106 CFU/mL. The dilution described here (100 μL of a 1.0 McFarland suspension in 10 mL of media) generates this inoculum for Klebsiella pneumoniae and other Enterobacteriaceae species on which we have tested it. If the density of the starting inocula in an experiment using different organisms is significantly higher or lower than this, then a different dilution may be needed. (The appropriate dilution required for a given species can be determined by performing a plate count of a 0.5 or 1.0 McFarland suspension to determine how many organisms this turbidity represents, then calculating the amount by which the initial suspension must be diluted to reach the appropriate final concentration.) If, on review of plate counts from the synergy study, the starting inoculum of any of the antibiotic-containing tubes is found to have been significantly lower than the starting inoculum of the growth control, this may indicate either antibiotic carryover or very rapid killing of bacteria in the brief time between addition of bacteria to the antibiotic-containing tube and removal of the aliquot for plating. If the actual number of colonies in the undiluted drop in a series is lower than the number of colonies in subsequent dilutions, this suggests antibiotic carryover effect. Different options have been described for preventing this effect, including spreading a single aliquot over an entire plate38 or spinning down the sample, removing the supernatant, and re-suspending in sterile saline prior to plating39. At each time point in the time-kill method, it is also critical for the investigator to efficiently but accurately remove an aliquot from each culture tube and perform serial dilutions. Delays during this process, particularly during early time points that occur in close succession, can lead to prolonged periods during which cultures are not been incubated and shaken, whereas careless dispensing and serial dilutions can lead to inaccurate plate counts. Compared to the spread plate method of plate counting, in which 100 μL of each dilution is spread over an entire agar plate, the drop plate method described is far more rapid, requires a much smaller number of agar plates, and allows for faster counting, as the maximum countable number of colonies for each drop is 30, whereas up to 300 colonies can typically be counted from a spread plate. However, the spread plate method is also an option if investigators are more comfortable with this technique. If drops spread into each other after dispensing with a multichannel pipette, individual application of more widely spaced drops with a single-channel pipette can be performed instead. In our experience, cooling plates at 4 °C prior to dispensing drops seemed to reduce excessive spreading.
One limitation of the techniques described here is that the results of the two types of synergy assay (checkerboard array and time-kill) are not always concordant, and since most published synergy articles use one method or the other rather than both together, it can be difficult to know how to integrate data from the two types of assays. Because the automated checkerboard array method we developed is simple and high-throughput, we have used it in effect as a kind of screen to test combinations against a larger number of isolates and to determine which concentration combinations were synergistic. We then performed a smaller number of time-kill studies, selecting combinations and concentrations that had been effective in the checkerboard array. Of note, because the checkerboard assay is typically performed on a microbroth dilution scale, while the time-kill assay uses larger volumes (similar to a macrobroth dilution), we found that FICs were sometimes different between the two methods, with higher concentrations generally required in the time-kill assay to demonstrate activity. This phenomenon has been noted previously when macrobroth and microbroth dilution MIC assay results are compared for Gram-negative bacilli26 and when larger inocula (as used in time-kill studies) are compared with the standard inoculum used in microbroth dilution and checkerboard array assays32. A specific limitation of the checkerboard array is the inherent variability in microbroth dilution MIC testing22. While FICI cutoffs for synergy account for this variability mathematically6 such variability inevitably raises concern about the reliability and consistency of checkerboard array results.
Because of the limitations inherent to all in vitro synergy testing methods (including cultivation of bacteria in an artificial growth medium, static antibiotic concentrations, and a limited time course), results obtained by these methods must be confirmed and further evaluated using supplemental techniques. Such methods include in vitro pharmacokinetic/pharmacodynamic (PK/PD) studies (e.g., the hollow fiber infection model40), animal models, and, ultimately, human PK/PD and efficacy studies. The automated checkerboard array method described here, by providing a rapid method with which to screen combinations for potential synergistic activity, allows for more targeted utilization of these techniques. Further automation of all of these methods, as well as more systematic investigation of the relationship between in vitro parameters and clinical outcomes, will be important in scaling up the use of synergy testing and increasing its clinical applicability.
The authors have nothing to disclose.
Thea Brennan-Krohn was supported by a Eunice Kennedy Shriver National Institute of Child Health and Human Development pediatric infectious diseases research training grant (T32HD055148), a National Institute of Allergy and Infectious Diseases training grant (T32AI007061), a Boston Children's Hospital Office of Faculty Development Faculty Career Development fellowship, and a National Institute of Allergy and Infectious Diseases career development award (1K08AI132716). J.E.K. was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award number R33 AI119114. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Escherichia coli strain ATCC 25922 | ATCC | 25922 | QC strain |
0.5 mL microcentrifuge tubes | USA Scientific | 1605-0000 | |
1 L 0.22 µm bottle-top filter | Thermo Scientific Nalgene | 597-4520 | |
12 mm x 75 mm borosilicate glass round bottom culture tubes | Fisherbrand | 14-961-26 | |
15 mL conical tubes | Phenix | SS-PH15 | |
15 x 100 mm or 15 x 150 mm Mueller Hinton agar plates | Thermo Scientific | R01620 or R04050 | |
25 mm stainless steel closures for 25 x 150 mm glass culture tubes | Bellco | 2005-02512 | |
25 x 150 mm borosilicate glass round bottom culture tubes | Bellco | 2011-25150 | |
348-well sterile clear, flat-bottom, untreated microplates with lids | Greiner Bio-One | 781186 | |
50 mL conical tubes | Phenix | SS-PH50 | |
50 mL sterile reagent reservoirs | Corning | 4870 | |
96 deep well polypropylene microplate with 2 mL wells | Fisherbrand | 12-566-612 | |
96-well sterile clear, round-bottom, untreated microplates with lids | Evergreen | 222-8032-01R | |
Cation adjusted Mueller Hinton broth | BD Diagnostics | 212322 | |
Colistin sulfate | Alfa Aesar | J60915 | |
D300e Control Software | HP/Tecan | ||
DensiCHEK Plus McFarland reader | bioMérieux | 21250 | |
Excel spreadsheet software | Microsoft | ||
Extra long SHARP 10 µL Precision Barrier Tips | Denville Scientific | P1096-FR | |
HP D300 digital dispenser | HP/Tecan | ||
HP D300 T8+ cassettes | HP/Tecan | 30097370 | |
Minocycline hydrochloride | Chem-Impex | 14302 | |
Picus 12-channel 10-300 µL pipette | Sartorius | 735461 | |
Polysorbate 20 | Fisher Bioreagents | BP-337 | Brand name: Tween 20 |
Sodium chloride | Fisher Chemical | S271 | |
Spectrophotometer | Tecan | Infinite M1000 PRO | |
Xplorer 12-channel 50-1200 µL pipette | Eppendorf | 2231000328 |