Detection and drug-susceptibility testing of M. tuberculosis from sputum samples using luciferase reporter phage: comparison with the Mycobacteria Growth Indicator Tube (MGIT) system

1. Introduction 

M. tuberculosis infects one-third of the world’s people (Raviglione et al., 1995), and while generally curable, is becoming increasingly resistant to commonly used antibiotics (Espinal et al. 2001; Pablos-Mendez et al., 1998). This resistance compromises the otherwise successful outcomes expected with implementation of the DOTS (directly observed short course therapy) strategy for tuberculosis (TB) control (Espinal et al. 2000). In fact, use of “standard” therapy in the face of unsuspected drug resistance can lead to nosocomial spread of TB (Edlin et al., 1992), and high mortality rates, especially among those HIV co-infected (Turett et al., 1995). These complications can be prevented by prompt diagnosis and institution of appropriate therapy (Turett et al., 1995). As HIV spreads into parts of Asia where M. tuberculosis infection is already endemic, the incidence of TB disease, and drug-resistant TB is expected to increase (Espinal et al. 2001). Thus, inexpensive and rapid tests to identify M. tuberculosis and the most appropriate medicines to use in therapy, will be highly desirable.

Detecting TB, and identifying drug-resistant TB by traditional microbiologic methods, is complicated by the slow growth rate of the causative agent, M. tuberculosis Heifets and Cangelosi 1999, Heifets and Good 1994. Therefore, surrogates of mycobacterial growth, such as oxygen consumption from the media, have been harnessed to develop optimized phenotypic detection methods such as the MGIT (Mycobacterial Growth Indicator Tube) system, which allows detection of M. tuberculosis in about 2 weeks (Alcaide et al. 2000). This system is also being developed for antibiotic susceptibility testing (AST), requiring a median of 8 days in a manual format (Rusch-Gerdes et al., 1999), and 6.5 to 9.5 days in an automated format Bemer et al 2002, Tortoli et al 2002, but it is not yet approved for this indication, and the technology is expensive. While molecular methods hold promise for rapid detection of genotypic correlates of drug-resistance, their utility is compromised by the multiple resistance mechanisms for most drugs (Riska et al. 2000), and the expense of these tests for use in the developing world where TB is concentrated. We have developed a rapid phenotypic identification and susceptibility test for M. tuberculosis which is simple to perform and relatively inexpensive. It relies on a recombinant mycobacteriophage –the luciferase reporter phage (LRP) Carriere et al 1997, Jacobs et al 1993, Riska et al 1999. The LRP is a virus which specifically infects mycobacteria, and has been engineered to signal the presence of viable mycobacteria by its rapid production of the sensitive reporter protein, firefly luciferase (Fflux). This luciferase in the presence of the exogenously added substrate, luciferin and endogenous ATP produces quantifiable light, which can be detected with a luminometer, or with slightly less sensitivity but lower cost, with polaroid film (Riska et al., 1999).

We have previously shown that the LRP can specifically detect M. tuberculosis (Riska et al., 1997), with a sensitivity comparable to acid-fast smear (Carriere et al., 1997), and can further provide accurate susceptibility information for the major anti-tuberculous drugs Riska et al 1999, Riska and Jacobs 1998. In this work, we aim to extend these findings by i.) developing a new LRP, phAE142 which uses a potent promoter to drive luciferase expression and enhances sensitivity of detection; ii.) demonstrating the ability of phAE142 to detect mycobacteria directly from processed sputum pellets; iii.)comparing LRP with MGIT technologies for time to detection and species identification of Mtb; and iv.)determining susceptibility results from primary sputum isolates of Mtb using LRP.

2. Materials and methods 

Mycobacterium smegmatis strains mc2155 Carriere et al 1997, Jacobs et al 1993, Riska and Jacobs 1998 and mc24502 (see below) used to propagate phages were grown in Middlebrook 7H9 media (Difco, Detroit, Mi) containing 0.2% glycerol, 0.5% bovine serum albumin (BSA), fraction V, 0.2% dextrose and 0.09% NaCl (designated MADC) with 0.05% Tween-80 and 150 μg/ml hygromycin (for mc24502).

Phage phAE142 was derived by replacing the promoter driving Fflux expression, from Phsp60 in phAE85 (Carriere et al.1997) to Pleft from phage L5 (Brown et al., 1997). Briefly, pYUB566, the kanamycin-resistant Fflux cosmid in phAE85 was removed by Not1 digest, and replaced with pYUB585 using a previously described in vitro packaging approach (Carriere et al., 1997). pYUB585 is an ampicillin-resistant cosmid derived from pKB15 (Brown et al., 1997) which has had the attP-int integration sequence and hygromycin-resistance gene deleted, and which carries Fflux under the PLeft promoter. Propagation of phAE142 in wild-type M. smegmatis leads to large-plaque variants of the phage which invariably contain deletions in the Fflux gene (unpublished data). Thus, phAE142 requires a special propagating strain of M. smegmatis (mc24502), containing an integrated hygromycin-resistance bearing plasmid (pSSB8) constitutively expressing gp71 from Phsp60; gp71 is the L5 repressor protein which turns off PLeft-driven Fflux expression (Brown et al., 1997). The phage is maintained at a titer of 1011 PFU/ml by propagating on mc24502 in MADC without hygromycin (to avoid carry-over into clinical samples), using standard protocols (Riska & Jacobs, Jr. 1998).

Sensitivity of detection is considered to be the minimum number of organisms (CFU) detectable by light production of at least 3 times background. In comparing phAE85 with phAE142, 0.5 mL of tenfold dilutions of their host strain (Mtb Erdman at 5 × 107 CFU/ml) were infected with equivalent quantities of phage (50 ul of 1011 PFU/ml). Light was measured in duplicate 25 μL aliquots after 1,2,3,4,5 and 6 h of infection, in a Lumac 2010a luminometer which injects 100 μL luciferin buffer (0.33 mM D-luciferin (Molecular Probes, Eugene, Or) in 0.05M sodium citrate, pH4.5), and utilizes 10s integration time.

Sputum pellets were obtained from 20 acid-fast-smear positive study protocol patients in Uganda prior to therapy (Wallis et al. 2000). The sputum was serially digested with N-acetyl L-cysteine, and then decontaminated with 2% sodium hydroxide-1.45% sodium citrate, as described (Wallis et al. 2000), and resuspended in 0.067M phosphate buffer, pH 6.8, with 0.5 mL aliquots frozen at –70°C prior to shipping to the US. Each specimen had BACTEC 460 susceptibility tests performed in a reference lab in Uganda, but the investigators were blinded to this data until after study completion. In addition, each specimen was titered before freezing (Wallis et al. 2000) and after at least 6 months, upon thawing prior to use (in the US).

Reagents and antibiotics (Isoniazid (INH), Rifampin (RIF), Streptomycin (STR), Ethambutol (EMB)) were obtained from Sigma (St. Louis, Mo.) unless stated. Antibiotics were reconstituted to 20× working concentrations (2,40,40,100 μg/ml for INH, RIF, STR, and EMB) in sterile water, giving final concentrations of 0.1, 2, 2 and 5 μg/ml. Rifampin was initially diluted with ethanol. For 13 samples, the final concentration of STR used was 6 μg/ml, and for EMB was 1.7 μg/ml, which subsequently were found to be comparable to the other concentrations used in this study. NAP (para-nitro alpha aminoacetyl β-hydroxy propiophenone) was kindly provided by Dr. Salman Siddiqi (Becton Dickinson (BD), Sparks, MD) and diluted in 70% ethanol to a 10× concentration of 75 μg/ml. The MGIT PANTA (BD) antibiotic supplement was utilized to supplement both MGIT and MADC cultures (see below), but was reconstituted in 3 mL sterile water (to 50× strength) instead of MGIT-growth supplement mix (BD), since the detergent in the latter was found to be deleterious to phage infection (unpublished data).

3. Detection Testing Protocol 

Each sputum pellet was thawed and distributed as shown in Fig. 1, as follows:

View Large Image Download to PowerPoint

Fig. 1. Scheme of testing of clinical specimen (dark circle). Horizontal thin arrow: time to detection; Horizontal thick arrow: time for NAP test and AST; Up arrow: LRP assay; Cloud-burst: positive LRP reaction; Plus sign: positive flagged by MGIT machine. Media used for each stage are indicated (MGIT, MADC,7H10 or Blood agar).

a.) 150μl into a MGIT 960 bottle (7 ml) supplemented with 700 μL MGIT- growth supplement and 160μl of 50X PANTA, and loaded onto a MGIT 960 machine for 6 weeks or until flagged as positive by machine software.

b.) 50μl into 24-well plates (Falcon 3047, BD, Lincoln Park, NJ) with 2 mL MADC media containing 0.5% glycerol, 0.1% casamino acids, and supplemented with 40 μL of 50X PANTA(BD). These plates were sealed with Para-Film, placed on a gently shaking platform in a 37°C incubator and were tested by LRP infection at Days 1 or 2, 3 or 4, 5 or 6, 7 to 10, 11 to 14, 15 to 18, 19 to 22, 23 to 26, 27 to 30 and 40 to 42 for the presence of detectable light. At each point, 100 μL was removed and added to 10 μL of phAE142. After 4-6 hrs, 25 μL aliquots were removed and emitted light was measured by the usual protocol (see above). If the relative light units (rlu) were greater than 50, which is 3 times background (usually 10-15 rlu), the sample was considered positive and set up for AST and NAP testing.

c.) 10 μL was plated on a Blood agar plate (BAP; BBL, Cockeysville,Md.) held at 37°C to look for contaminants,

d.) 10 μL was serially diluted in MADC with 0.05% Tween-80 and plated on Middlebrook 7H10 plates (Difco) supplemented with 10% OADC (BBL) to assess viable counts (CFU/ml).

4. TB complex Assignment and Antibiotic Susceptibility test (AST) Protocol 

Mycobacteria from either LRP-positive MADC cultures or machine-positive MGIT tubes, off-loaded on a 5-day week schedule were used. MADC cultures were used directly, while 1.5 mL of MGIT media was centrifuged (5000 rpm, 5 min, room temperature) and resuspended in 600 μL MADC. Then, 100 μL aliquots of the culture were added to 5 μL of each antibiotic, or 10 μL NAP, or sterile water in 96 well plates. After 48 h. incubation, 5 μL of phAE142 was added to each well and light output was determined as described above in 25 μL aliquots. A viability index for each antibiotic was defined as: 100*(rluantibiotic –rlubckg)/(rluno_drug –rlubckg), where each rlu value is the mean of duplicate readings, and rlubckg is the rlu from media alone, generally 10-15 rlu. Viability index of <10% was considered the criterion for defining a strain as susceptible, and viability index of <25% after NAP assigned a strain to the M. tuberculosis complex (Riska et al., 1997). Certain discrepant strains had susceptibilities performed at our clinical laboratory with the BACTEC 460, using the manufacturer’s protocols (Siddiqi 1995).

Statistical analysis: A paired two-tailed t test in the Microsoft Excel program was used for comparisons of time to results.

5. Results 

5.1. Sensitivity of detection of M. tuberculosis 

In preliminary work with spiked sputum samples or clinical specimens using phAE85, it became apparent that M. tuberculosis could be detected in conventionally processed sputum immediately, but only if the starting inoculum had “numerous” acid fast bacilli, or > 107 CFU/ml (Hobby et al., 1973). However, by allowing a brief period of growth (1-7 days) in culture, light could be detected at a rate faster than the doubling time of the organism, suggesting improved infectability in the appropriate milieu. One approach to minimize this interval was to optimize the sensitivity of our LRP, and this effort has led to the development of phage phAE142. In head-to-head comparisons with phAE85 at comparable titers, phAE142 produced 10 to 100-fold more light from a given concentration of M. tuberculosis Erdman strain (Fig. 2). (Table 1) This yields an analytical sensitivity, using a 3-times-background threshold of detection of 1300 CFU per measurement (on repeat testing: 400, 1300, 1300, 2000 CFU), which is unprecedented compared to our other LRP constructs used with M. tuberculosis. This phage was therefore applied to primary sputum specimens in this study.

View Large Image Download to PowerPoint

Fig. 2. Sensitivity of detection using phAE142 (A) vs. phAE85(B), both at 1011 PFU/ml, tested against tenfold dilutions of M. tuberculosis Erdman, starting with 5 × 107 CFU/ml through 5 × 103 CFU/ml, as indicated in the legend. Each measurement represents light (rlu) from 25 μL of the mycobacterial suspension; background light is 10 rlu.

Table 1. AST results of M. tuberculosis strains assessable in either MADC-LRP or MGIT-LRP assays, comparing original BACTEC 460 AST result with LRP AST for the same isolates.

	Antibiotic	Susceptibility Result by		Cohort of strains evaluable by
		BACTEC	LRP	MADC-LRP- AST (n = 17)	MGIT-LRP- AST (n = 19)
	INH	S	S	12	14
		R	R	2	3
		S	R	2	1
		R	S	1*	1*
	RIF	S	S	16	18
		R	R	1	1
	STR	S	S	15	17
		R	R	1	1
		R	S	1	1
	EMB	S	S	17	19

S: Susceptible; R: Resistant

* discrepant stain found on BACTEC re-testing to be S to INH (concordant with the LRP result)

6. Detection of M. tuberculosis in clinical samples:phae142 compared to mgit 

Twenty freshly thawed processed tuberculous sputum pellets were titered on 7H10 plates, with the range of initial CFU/ml shown in Fig. 3A. One specimen failed to grow on 7H10 (limits of detection:100 CFU/ml), and no microbial contamination was seen. In contrast, one processed sputum pellet was initially contaminated, with heavy non-mycobacterial growth on BAP. Despite use of a common PANTA antibiotic stock, this contaminant was suppressed in MGIT media, allowing prompt detection, but it did compromise mycobacterial growth in the MADC media, leading to late detection and preventing performance of AST by the LRP assay.

The results of our time-to-positivity trial, comparing MGIT 960 automated detection to LRP detection are stratified by the initial CFU/ml as seen on 7H10 plates (Fig. 3A) . or compared head-to-head (Fig. 3B). While MGIT tubes all became positive (median 9 days, range 2-12), 1 of 20 specimens was turbid and was found to be contaminated with non-mycobacterial growth (false positive). In comparison, 18 of 20 specimens yielded detectable mycobacteria in the MADC-media by LRP assay (median 7 days, range 4-26)–one specimen with delayed detection was found to be contaminated (see above). In fact, the 2 specimens which failed to grow in MADC media had the lowest titers, one with no growth on 7H10 and the other well below 104 CFU/ml, which would be considered acid-fast smear-negative had these been fresh specimens (Hobby et al., 1973). On the other hand, of the specimens with highest initial CFU/ml (>105 CFU/ml), 8 of 9 were positive by LRP as soon as or earlier than MGIT, by a median of 2 days (p = 0.15).

View Large Image Download to PowerPoint

Fig. 3. A.) Time to detection of mycobacteria by MGIT or LRP, in days, stratified by initial titer of mycobacteria (CFU/ml) in the inoculum. Time to a positive MGIT by machine detection (square), linked by solid line to the time of a positive LRP assay (rlu > 50) performed on the same specimen (triangle). Time to a positive LRP assay from specimens grown in MADC(circles). B.) Time to detection comparing growth in MGIT (machine-positive) with growth in MADC and LRP detection (rlu > 50). Line of equivalence is depicted, with faster results with LRP shown to the right, faster results by MGIT to the left.

7. LRP phAE142 detection of M. tuberculosis from machine-positive mgit bottles 

Preliminary data revealed that M. tuberculosis could be detected in positive MGIT bottles by LRP, especially if the specimen is washed and resuspended in MADC, thereby removing the detergent components in the MGIT media. Generally, optimal light is obtained only when LRP infection is performed the next day after washing, so we chose to wash and set up AST tests immediately upon detecting a MGIT 960 positive bottle, without prior screening for phage infectability. In practice, this meant the first LRP assay from MGIT was performed a median of 4 days (range 2-8 days) after machine-positivity (Fig. 3A), at which time AST and species data were also available (see below). The variability was largely determined by the day of set-up for AST, as all initial LRP assays were positive (except the 1 clearly contaminated). A surprising observation, although consistent with the higher isolation rate in MGIT vs. MADC media, was that the light produced from mycobacteria grown in MGIT media was significantly higher than that from the same isolates grown in MADC- the light units obtained per day of growth were a median of 1 log higher, and were higher in 15 of 17 strains which yielded light in both media (p = 0.0002). Initial titers (CFU/ml) in these cultures had no bearing on the growth rates observed, while the increased light output corresponded to increased numbers of bacilli, at the time of detection, where assessed (data not shown). In subsequent experiments, this growth rate advantage of MGIT media was found to hold for another 2 primary sputum isolates, but not for subcultured isolates, suggesting that the MGIT media is most favorable for nurturing those isolates obtained from the clinical environment, and/or subjected to harsh processing (freeze/thawing, alkaline decontamination).

8. LRP phAE142 used to assign isolates to the TB complex 

In our clinical mycobacteriology laboratory, as in others (Alcaide et al. 2000), a positive MGIT bottle is confirmed to contain M. tuberculosis by the presence of cording on AFB smear, followed by a DNA-probe assay performed a median of 3 days later. By comparison, any clinical isolate producing light after LRP infection should be strongly considered to be M. tuberculosis, since other clinical mycobacteria strains are less efficiently infected by the TM4 phage family (Riska et al., 1997). Further, positive LRP assays are further confirmed to result from M. tuberculosis complex strains by the subsequent inhibition of light output (>75%) in the LRP format of the NAP assay, which could be available within 48 h of detection (Riska et al., 1997). In this trial, the results were obtained a median of 3 days after detection in the MADC format, and 4 days in the MGIT format. Overall, 17 of 17 specimens tested from MADC were inhibited by NAP, as were 18 of 19 from MGIT, for an aggregate sensitivity of 97%. The one discrepant strain was repeated and was now clearly inhibited by NAP.

9. LRP phAE142 used for AST of primary clinical specimens and positive MGIT bottles 

Once light was obtained from an MADC culture using LRP, or a MGIT bottle became positive, an AST assay for the 4 conventionally tested antibiotics was set up. The overall time to AST results tended to be shorter for the MADC cultures than for the MGIT bottles in most instances (with a median of 11 days for MADC cultures, vs. 13 days for MGIT (t test, p > 0.5). Nonetheless, all the noncontaminated MGIT cultures had AST results by 16 days, whereas some of the MADC cultures took longer for AST results. Although standard AST using BACTEC was not repeated for these isolates, it would be expected to take a median of 2 weeks once an isolate was detected in MGIT in our clinical laboratory (M. Levi, personal communication), or 3 to 4 weeks from specimen acquisition, within the guidelines recommended by the CDC (Bird et al., 1996).

While most of our isolates were drug susceptible, there were 3 INH-resistant, 1 STR-resistant and 1 INH, RIF and STR-resistant strains defined by initial BACTEC testing. The accuracy of the AST results performed by LRP is depicted in the Table, with isolates grown in MADC and MGIT media considered separately. There is 100% agreement seen between methods for RIF and EMB, 86% for INH, 94% for STR and 95% overall.

Two discrepant strains gave identical results when grown in either MADC or MGIT media and tested with LRP. These were 1) an initially INH-resistant strain which upon repeat testing in our clinical laboratory by BACTEC 460 was found to be INH-susceptible at 0.1 μg/ml; and 2) a streptomycin-resistant strain (confirmed by BACTEC 460 using 2 μg/ml); which was STR “susceptible” on repeat LRP testing using a concentration of 2 μg/ml, though “resistant” at 0.5 μg/ml, only 1 dilution higher than our reference susceptible strain. If the first strain were counted as INH-susceptible, then accuracy for INH is improved to 92%, and overall it would be 97%.

Three other discrepant strains (2 from MADC, 1 from MGIT) were falsely interpreted as INH resistant, inasmuch as they were incompletely inhibited by INH 0.1 μg/ml, with viability indices between 20 and 40%. In contrast, the 3 truly resistant strains had indices of nearly 100%. As suggested by others (Banaiee et al. 2001), a breakpoint for INH of 0.2 μg/ml should be considered. One false-resistant isolate re-tested by LRP was completely inhibited by INH 0.2 μg/ml, and confirmed to be INH-susceptible by BACTEC; another was inhibited by INH 0.1 μg/ml on re-testing and the third was again resistant at 0.1 μg/ml and not available for testing at 0.2 μg/ml.

10. Discussion 

Early detection of M. tuberculosis and its drug susceptibility pattern is being increasingly recognized as an important component of global TB control (Espinal et al. 2000). We here describe a protocol using luciferase reporter phages which can perform this task in 1 to 2 weeks from primary specimens in a format adaptable to the developing world where TB is so prominent.

To optimize this system, we utilized the most sensitive phage developed to date for the detection of M. tuberculosis, phAE142, which features a promoter (Pleft of heteroimmune mycobacteriophage L5 (Brown et al., 1997)) which seemed to be more potent than the previously utilized Phsp60. When used to drive luciferase expression on comparable integrating vectors incorporated into M. smegmatis, Pleft (in mc24501) produced 100-fold more luciferase than Phsp60 (in mc24503) at comparable cell densities (data not shown). Further, Pleft drove luciferase expression to toxic levels in both multi-copy vectors (Brown et al., 1997) and in the phage construction process –it was thus only after the promoter was silenced by its repressor that a phage carrying it could be stably propagated. In addition, we explored nearly 20 alternate sputum processing protocols (including all those in clinical use (Kent & Kubica 1985)) (data not shown) before concluding that the preferred NALC-NaOH method, optimized for other phenotypic detection methods, is also most appropriate for LRP. As with other liquid culture systems for mycobacteria, selective antibiotics (PANTA) were required to minimize contaminants, which occurred within the acceptable range of less than 5% Alcaide et al 2000, Siddiqi 1995; Tortoli et al. 2002). It is possible though unlikely that the freeze-thawing might have lowered our contamination rate.

For detection of M. tuberculosis, the performance of sequential LRP testing of primary cultures was compared to the MGIT system with continuous-reading hardware. LRP had earlier detection at higher inocula, while MGIT proved to be more reliable with inocula less than 104 CFU/ml. The comparison was biased in favor of MGIT by virtue of the latter’s continuous reading schedule, compared to the intermittent readings for LRP; and the time-to-positive for MGIT also failed to include the time for performing a confirmatory AFB smear, which on average required an additional day. However, samples grown in MGIT medium consistently yielded more light after LRP infection, not merely due to the twofold concentration afforded by centrifugation but correlating with improved growth of primary cultures of M. tuberculosis in this media. The mechanism of enhanced growth is being investigated, although it likely is due to the MGIT media formulation (more 7H9, presence of oleic acid and detergents) compared with MADC. Interestingly, 2 low-titer isolates grew in MGIT but not MADC, perhaps reflecting the higher volumes inoculated into MGIT in this trial.

Several caveats should be noted in comparing our time to results by LRP with the MGIT system. First, we were constrained by the available volumes of sputum pellets to inoculate 150 μL into MGIT, instead of the recommended 500 μL, potentially delaying detection. However, we matched our MADC inoculum to provide the same final concentration of mycobacteria as the MGIT inoculum, and a parallel study (Banaiee et al. 2001) which did use 0.5 mL, showed similar results to ours.

Second, we have restricted our analysis to smear-positive specimens. Nonetheless, the fact that these were freeze-thawed had decreased their viable titers so that they represent a range of different mycobacterial inocula (Fig. 1). Had we used fresh smear-positive specimens, the initial titers of mycobacteria would be higher, and detection times could be even shorter. Finally, although negative specimens were not included here, false-positive detection of mycobacteria is nearly impossible with LRP given the exceptional specificity of our phages for mycobacteria, and the published effects of NAP in allowing definitive recognition of TB complex organisms (Riska et al., 1997).

Third, our LRP detection protocol is labor-intensive in its present format. In the clinical laboratory, LRP readings would probably be performed at limited intervals such as 2 to 3 days after inoculation, and then weekly thereafter. These intervals may increase turn-around time, but this may be acceptable if the alternative is no AST; they will need to be optimized in future studies. Alternatively, examination for microcolonies on thin plates (Mejia et al., 1999), or in microwell liquid cultures (Caviedes et al. 2000), could be the primary, low-cost screening method for identifying positive cultures, followed up by LRP for species confirmation and AST. Finally, in assessing times for AST and species assignment, we have compared our LRP results from this research study with “averaged” data from our clinical laboratory, which incorporates delays due to batching specimens for testing. Still, our results are 50% faster than the target of 28 days for AST results, which had been met by only a minority of state mycobacteriology laboratories despite their common (80%) use of BACTEC methods (Bird et al., 1996).

The LRP technology is particularly suited to performing AST, and is the most rapid phenotypic method. It is specific for M. tuberculosis (especially in conjunction with NAP), so less vulnerable to non-specific results due to contaminants or co-infections. Finally, it is simple to perform, provides objective data, has potential safety feature in that phage ultimately lyses infected mycobacteria and has relatively low reagent cost. Its accuracy in this work, though limited due to the small number of resistant isolates, is comparable to other studies using LRP Banaiee et al 2001, Riska et al 1999.

While discrepant analysis has inherent biases (Hadgu, 2000) and was not used to affect our overall accuracy, it does lead to hypotheses which can be tested in future studies. One strain supposedly INH-resistant repeatedly tested susceptible in our hands, perhaps reflecting an artifact from a mixed-resistant population in the original specimen, or a strain with true borderline susceptibility. More intermediate degrees of inhibition of light output (between 50 and 90%, rather than > 90%) were seen with INH in several cases, possibly due to the effects of sputum components carried over which might partially inactivate INH –testing higher concentrations of INH is therefore warranted in this setting, and may clarify this issue. Finally, one streptomycin-resistant strain was repeatedly misidentified by LRP. Lowering of the testing concentration to 0.5 μg/ml would have correctly identified this isolate, and was evaluated in a larger, parallel study (Banaiee et al. 2001) with good results. In any case, given the short turnaround time of this assay, suspicious results, or any resistance to antibiotics or NAP, can be readily re-tested and confirmed.

In summary, we describe a more sensitive luciferase reporter phage assay utilizing phAE142, which can be applied to primary sputum specimens grown until positive in either phage-testing media (MADC) or MGIT media. Based on the sensitivity of the phage, this test requires a sample with 0.5 to 1 × 105 CFU/ml, at which time rapid and reliable AST and species confirmation can be set up in a simple microwell format. In this study, the test results were available at least 1 week before those reported for any available phenotypic susceptibility assay. However, the specimens used had been frozen for at least 1 year before use, suggesting that fresh specimens might have results even earlier. Further trials utilizing direct specimens from regions with higher background resistance rates are warranted, as are efforts to optimize sputum processing and outgrowth. The most sensitive assay would be desirable, in that it may allow use of the “low-tech” photographic detection modality (Riska et al., 1999) which would be most useful in the developing countries where TB is most prevalent.