|Home | About | Journals | Submit | Contact Us | Français|
Methods currently used for in vitro drug susceptibility testing are based on the assessment of bacterial growth-related processes. This reliance on cellular reproduction leads to prolonged incubation times, particularly for slowly growing organisms such as mycobacteria. A new rapid phenotypic method for the drug susceptibility testing of mycobacteria is described. The method is based on the detection of the physiological stress developed by susceptible mycobacterial cells in the presence of an antimicrobial compound. The induced stress was quantified by differential monitoring of the dielectric properties of the bacterial suspension, an easily measurable electronic property. The data presented here characterize the stress developed by Mycobacterium tuberculosis cells treated with rifampin (rifampicin), isoniazid, ethambutol, and pyrazinamide. Changes in the dielectric-based profiles of the drug-treated bacteria revealed the respective susceptibilities in near real time, and the susceptibilities were well correlated with conventional susceptibility test data.
Drug-resistant tuberculosis (TB) is of increasing clinical and financial concern for health care systems worldwide (14, 21). Absent or delayed detection of drug resistance is one of the major contributors to outbreaks of multidrug-resistant and extremely drug-resistant TB (16, 24). Consequently, drug susceptibility testing (DST) is essential for the initiation of therapies with properly adjusted regimens and surveillance of the susceptibility patterns of the strains circulating within the population.
Currently, all nonmolecular in vitro drug susceptibility measurements rely exclusively on the detection of bacterial growth (cell division) or growth-related metabolites either in liquid medium or on solid medium to obtain accurate and reliable test results. In this report, we describe a new, rapid approach for determining the drug susceptibility of mycobacteria without detecting growth-specific metabolites or cell proliferation.
It is well documented that some chemical compounds (including antimicrobial drugs) can trigger a measurable stress response in susceptible bacterial cells (8, 28). The intensity of the response is also a measure of the effect of a drug on the organism. Because the activation of stress does not depend on the growth rate, DST results can be obtained in near real time with the availability of an appropriate sensing modality.
We have found that monitoring the dielectric permittivity of a biological suspension is a simple and practical method for characterization of the development of the physiological stress response and the associated ancillary processes of the bacteria. The dielectric permittivity reflects changes in the apparent charge distribution within the suspension and is an easily measurable electronic property obtained from measurements of the capacitance of a sample. The inclusion of differential sensing methods enables the isolation of the stress-specific signals from other unrelated signals within the system.
As cellular anabolism/catabolism progresses, the dielectric permittivity is modified, reflecting changes in cell morphology and alterations in redox/membrane potentials, surface charge effects, and synthesis, as well as changes in the conformations of highly charged molecules (DNA and RNA). This includes modifications to the compositions of salts, proteins, amino acids, and other constituents of the medium. The signal detected by our system is the cumulative response from all metabolic processes and, therefore, reflects phenotypic changes in the responding cells. It is important to note that the detection of dielectric changes during the development of stress are not related to changes in cell numbers.
To illustrate the potential of this new approach, data are presented for the model strain Mycobacterium tuberculosis H37Ra exposed to the four first-line anti-TB drugs rifampin (RIF; rifampicin), isoniazid (INH), ethambutol (EMB), and pyrazinamide (PZA). These four drugs are the basis of the currently recommended anti-TB therapy, and DST with the last two drugs is known to be challenging (2, 20).
All measurements of the dielectric permittivity were made by using instrumentation custom-built for these experiments. The measurement system consisted of a cassette that holds the test samples and a temperature-controlled tabletop-sized device into which the cassette is inserted for analysis. Photographs of the apparatus and a test cassette are shown in Fig. Fig.1.1. All cassettes used in these experiments contained two 100-μl test chambers, with each having the same thermomechanical properties for optimal differential measurement. Both chambers were defined by a submillimeter-sized electrode gap structure that encompassed a sample and sensed its respective electrical properties. The electrode surfaces were made of pure gold deposited onto a glass plate (Borofloat; Schott Glass, AG).
The capacitance of each sample was obtained directly from a measurement of the electrical impedance of the suspension and was used to calculate the dielectric permittivity. The impedances for each detection chamber were continuously measured over time and stored electronically in a personal computer. The respective impedances were then compared to minimize any changes to the dielectric properties common to both chambers. These common properties include nonessential contributions to the impedance that can obscure the signal from bacteria in the complex biological system.
For the planar geometry used in the design of these cassettes, the dielectric permittivity is directly proportional to the measured capacitance. For all experiments, the dielectric permittivity was calculated by the equation C = (A/d), where C is capacitance, is the dielectric permittivity of the suspension, A is the electrode area, and d is the separation distance between the electrodes. A normalized impedance response (NIR) was calculated by dividing the signal value from the chamber containing the drug-treated cells by that from the reference (control) chamber containing the same type of cells not treated with drug and is presented in relative units. The NIR is a quantity that minimizes elements of the signal not directly associated with the effects of the stressor on viable cells. The values of the NIR profiles were arbitrarily scaled to equal 1.000 at the start of each experiment for ease of comparison. Otherwise, no other modifications to the NIR values were made.
All data presented in this report were obtained by using Mycobacterium tuberculosis H37Ra (ATCC 25177; ATCC, Manassas, VA). This strain was susceptible to all of the drugs tested and served as the parental strain for all resistant mutants developed in-house. The four drugs used in this research, RIF, INH, EMB, and PZA, were purchased from Sigma (St. Louis, MO).
All strains were grown to the mid- or late-logarithmic stage of growth at 37°C in Middlebrook 7H9 broth (Sigma) or on Middlebrook 7H10 agar (Sigma) with the recommended supplements.
Sputum samples obtained at a local hospital from anonymous patients who had been demonstrated to have etiological lung pathogens other than acid-fast bacilli were spiked with 107 CFU/ml of the model organism. The sputum was then processed with 1% NaOH and 2% N-acetyl-l-cysteine, according to the protocol of the WHO (27). The mycobacterial cells were collected by centrifugation, resuspended in 7H9 broth with supplements, and allowed to recover before further experimentation. Cell numbers and viability were confirmed before and after treatment by standard plating methods and acridine orange staining. Typically, 70% of the spiked cells survived the sputum processing.
An important part of this investigation was recording the changes in the dielectric permittivity of mycobacterial strains resistant to the drugs used in our study. Consequently, spontaneous mutants of M. tuberculosis H37Ra resistant to RIF, INH, EMB, and PZA were successfully established. The resistant mutants were isolated as described by Kosagöz and coauthors (10). Briefly, the parental strain in mid-logarithmic phase was serially propagated alternately in 7H9 broth and then on 7H10 agar plates containing progressively increasing concentrations starting from 0.5× MIC of the respective drug. Several mutants resistant to each drug were obtained; the one with the highest level of resistance was used to obtain the impedance profiles.
All antimycobacterial agents used were tested before any impedance measurements were conducted by a standard broth microdilution method, as described previously (26). Briefly, 106 CFU/ml was inoculated in 7H9 complete medium supplemented with twofold drug dilutions. After the recommended 21 days of incubation at 37°C, the cultures were visually inspected for evidence of turbidity, and the turbidity was compared to that of control cultures to which no drug was added. The corresponding MIC was determined as the concentration in the well with the lowest drug concentrations and no visible growth.
All DST by impedance sensing was conducted by using the following protocol. At 24 h after sputum treatment, the impedance response of the cells was monitored and compared with that of untreated cells to determine the extent of the recovery from alkaline stress. The next set of experiments was begun if the two respective impedance responses showed no statistically significant differences between each other, which inferred no stress response and, hence, the absence of effects from the treatment. Typically, alkaline-treated cells required between 24 and 48 h after sputum treatment for recovery. Once recovered cells were available, the resulting suspension was diluted in fresh 7H9 broth with or without an antimicrobial to approximately 105 CFU/ml, the concentration used for DST in many mycobacteriology laboratories (22). One chamber of the cassette was manually filled with a suspension containing either drug-treated or non-drug-treated mycobacteria, while the adjacent reference chamber was filled with non-drug-treated cells for direct comparison. The filled cassette was then inserted into the analyzer platform set at 37°C, and the impedance signals from each chamber were continuously recorded and analyzed. All experiments were repeated at least three times.
For measurements involving PZA, the pH of the 7H9 medium was adjusted from its original value of 7.0 down to 5.5 to enhance the activity of the drug. All other aspects of these experiments were carried out in a manner identical to that for the other three drugs.
The MICs for the susceptible parental mycobacterial strain for each drug tested are summarized in Table Table1,1, along with published values for similar measurements. The MICs measured for this strain are in line with the previously published values. We note that the published MICs in references 3, 19, 25, and 26 were determined with M. tuberculosis H37Rv or clinical isolates, while M. tuberculosis H37Ra was used for determination of the MIC published in reference 29.
The MICs for the mutant strains selected for their drug resistance are summarized in Table Table2.2. The corresponding MICs of the drugs with which mutants were selected were more than 80 times higher than the MICs for their corresponding parental susceptible strains. However, these mutant strains continued to maintain the same susceptibilities as their parental strain to the other three drugs.
The NIR profiles of both susceptible and resistant strains of M. tuberculosis treated with RIF, INH, EMB, and PZA were obtained; and the results are presented as composite plots in Fig. Fig.2.2. To investigate the dependence of the dielectric characteristics of the treated mycobacterial suspension on the drug concentration, we obtained the NIR profiles of the susceptible M. tuberculosis cells incubated in 7H9 broth in the presence of no drug, 2× MIC, and 10× MIC for all four drugs (Fig. (Fig.2,2, left panels). NIR profiles were also obtained for the corresponding drug-resistant strains exposed to a high concentration (10× MIC for the susceptible strain) of the drug. As a positive control, additional profiles were also recorded for drugs to which these strains were susceptible. The results of these experiments are plotted in Fig. Fig.22 (right panels) adjacent to the plots for the respective susceptible strains and with the same scaling to allow direct comparison of the differences between the NIR profiles of susceptible and resistant cells.
The NIR values of the bacterial suspensions to which no drugs were added (labeled untreated) showed no significant deviation from a constant value. This result was expected, since the respective signals from each chamber were generated by identical untreated cell suspensions. These profiles served as negative controls to which all other profiles of drug-treated cells should be compared. Overall, all the NIR profiles obtained for M. tuberculosis H37Ra cells treated with each of the four first-line drugs were qualitatively similar, irrespective of the mechanisms of action of the drugs. Specifically, a continuously decreasing NIR value was observed for cells susceptible to the antimicrobial, whereas a constant NIR value similar to the response when no drug was added was detected for cells resistant to the antimicrobial. The NIR profile for all susceptible cells was easily differentiated from that for the untreated control cells and/or the drug-treated resistant strain by visual inspection in near real time.
In addition, all NIR profiles of the susceptible strain showed a more pronounced response when the concentration of the drug was increased. For any time point, the changes in NIR values for cells incubated with 10× MIC of the drug (blue curves) were greater than the corresponding changes for cells incubated with 2× MIC of the same drug (orange curves). While incubation of the mycobacteria with drugs produced NIR profiles with negative slopes, the incubation was not associated with a detectable loss of cell numbers, confirming the presence of stressed but viable cells. As an additional control, the NIR profile was obtained when no cells were added to the medium to determine if the presence of the drug per se affected the baseline values of the NIR. For these experiments, one chamber was filled with 7H9 medium supplemented with the highest concentration of the drug, and the adjacent chamber contained the 7H9 medium alone. In all cases, the respective NIR values were constant. Thus, all NIR profiles plotted in Fig. Fig.22 reflect the effects of the anti-TB drugs on the mycobacterial cells without bias.
In summary, the responses of the different drug-susceptible mycobacteria combinations that were observed were qualitatively similar and shared one prominent feature: a nearly immediate and continuous decrease in the NIR value at an intensity proportional to the drug concentration used. To date, all experiments conducted in our laboratory with 14 different gram-negative or gram-positive species and 12 different antibiotics revealed similar NIR profiles and dependencies, suggesting the universality of the approach (17, 18).
The statistical significance of the NIR profiles was analyzed for the susceptible strain treated with each of the four drugs tested and its corresponding drug-resistant strain. Because the NIR profiles were measured continuously, the average values of the NIR at a single time point (60 min) were calculated along with the respective P values by using the t test. The results for the drug PZA are shown graphically in Fig. Fig.3.3. Similar comparisons for the other three drugs also yielded statistically significant results.
Bacterial cells respond to adverse environmental conditions by rapidly altering different cellular processes (cell division, membrane composition and potential, nucleic acid metabolism, housekeeping protein expression, etc.) (1, 6, 7). Functionally, this stress response represents metabolic events distinct from processes associated with cell division (5, 12). Because antibacterial drugs induce the stress response in susceptible cells, monitoring of the development of this stress is an attractive tool for bacterial DST. However, all currently existing methods capable of detecting the stress response (Western blotting, proteomic and transcriptomic analysis, Southern blotting, PCR, and mass spectrometry analysis) are impractical for most clinical laboratories. These molecular-based techniques require cell disruption, are unable to monitor the response continuously, are technically complex, and can detect only certain stress-related macromolecules. The impedance-based method for continuous monitoring of the development of stress reported here is noninvasive, technically simple, and operationally easy to implement. Thus, this approach holds promise for being developed as a cost-effective method for rapid DST, especially in developing countries where the need is great.
Measurements of bacterial growth (or the lack thereof) are the conventionally accepted approaches for determining the susceptibility of a bacterial strain to a drug. By these methods, the bacterial strain is judged to be either susceptible or resistant on the basis of a comparison of the lowest concentration of the antimicrobial needed to prevent the multiplication of the bacteria or to inhibit the occurrence of specific metabolic by-products. The MIC is considered the standard for all drug susceptibility measurements. An alternative way to characterize the effect of a drug on a bacterial strain is to measure complementary information that reflects the consequential metabolic status of the bacterial cells.
The approach investigated in the study described here accomplishes this by monitoring the physiological stress developed during exposure to a chemical stressor, such as an antimicrobial drug, in the mycobacterial cell. For example, an untreated susceptible strain maintained under optimal growth conditions will develop no stress, while the same cells exposed to a lethal concentration of a drug will be highly stressed as the cells seek to survive. Similarly, a fully resistant strain will show no stress at all during exposure to a drug concentration lethal to susceptible cells. The NIR profiles presented here for both susceptible and resistant strains of M. tuberculosis H37Ra treated with RIF, INH, EMB, and PZA conform to this theoretical model and form a complete set of susceptibility data. For all susceptible strains, the development of a measurable response was easily recognizable by visual inspection in less than 60 min, irrespective of the mechanism of action of the drugs or any differences in the growth rate of the respective strain.
Data were presented for mycobacterial cells treated with the drugs at concentrations at multiples of the MIC of the susceptible strain. The intensity of the NIR changes varies with the drug concentration used, suggesting the development of proportionate levels of stress within a susceptible cell. This conclusion is supported by the results published by Provvedi et al. (15), who found that the number of genes activated in mycobacteria treated with vancomycin increased as the concentration of the drug was increased. Conversely, we have found that as the concentration of the drug is reduced below MIC levels, more time is required for the corresponding NIR profile to develop a response that deviates from that for the negative control. The sensitivity of the NIR profiles to the drug concentrations suggests the possibility that an algorithm may be used to connect the NIR values to conventional MICs.
Of particular interest was the testing of PZA. Conventional DST of PZA is known to be inconsistent and challenging because of the poor growth conditions for mycobacteria at the low pH which is necessary to activate the drug (4, 11, 13). PZA is especially important for TB control programs employing directly observed therapy-plus guidelines for the treatment of multidrug-resistant TB. Its inclusion has dramatically reduced treatment times to about 6 months, from about the 9 months of treatment required without PZA and from the 24 months of treatment required under the INH-streptomycin-para-amino salicylate regimen used in the 1970s (9). Consequently, there is a great need for a rapid test to determine if this first-line drug is clinically suitable for use for the treatment of a given case. The rapid selection of adequate combinations of first-line drugs is further necessitated by complications of drug interactions presented in patients coinfected with human immunodeficiency virus and M. tuberculosis undergoing aggressive antiretroviral therapy (23). The enhanced sensitivity of the differential impedance-sensing modality, in combination with its relative insensitivity to pH, provides a unique means of evaluation of mycobacterial susceptibility to PZA.
Here, we present a simple, rapid, sensitive, and versatile approach for determining the drug susceptibility of mycobacteria. Thus, our results demonstrate that monitoring the stress developed by mycobacterial cells can be used as a truly rapid means to distinguish between drug-susceptible and -resistant strains. Future experiments are being planned to test the performance and robustness of our method with clinical isolates of M. tuberculosis and to compare our approach with “gold standard” culture-based methods.
This work was supported by funds from the National Science Foundation under grant OII-0750054, which are gratefully acknowledged.
We thank Eric Rubin and Natalia Kurepina for their very helpful comments in the preparation of the manuscript.
R. J. Rieder, Z. Zhao, and B. Zavizion are all employees of BioSense Technologies, Inc.; R. J. Rieder and B. Zavizion hold ownership positions in the company.
Published ahead of print on 24 August 2009.