Rapid Tuberculosis Diagnosis Using Reporter Enzyme Fluorescence

Tuberculosis is the most frequent cause of death in humans from a single infectious agent. Due to low numbers of bacteria present in sputum during early infection, diagnosis does not usually occur until >3 to 4 months after symptoms develop. We created a new more sensitive diagnostic that can be carried out in 10 min with no processing or technical expertise.

and smear microscopy. We evaluated conservation of BlaC in tuberculosis complex bacteria and compared it to ␤-lactamases present in other bacterial species. We synthesized the BlaC-specific REF substrate CDG-3 and found that it has high selectivity for M. tuberculosis BlaC and can quantitatively measure enzyme and bacterial levels in sputum. These characteristics allowed development of a simple sputum-based diagnostic assay for tuberculosis that can now be evaluated in patients (48)(49)(50). In the current study, we optimized conditions for REFtb to diagnose tuberculosis using sputum and conducted the first blinded trial of REFtb in patients with clinical symptoms of tuberculosis. We found that, based on the sensitivity and specificity obtained using BlaC-based REF for diagnosis of tuberculosis, REFtb represents a promising new triage test for tuberculosis that could be used to improve case identification prior to follow up by appropriate treatment and confirmation with other diagnostic strategies.
quences of ␤-lactamases were aligned and then a phylogenetic tree was constructed (58). A simple unrooted tree depicting relatedness of ␤-lactamases was generated.
Bacterial strains and growth conditions. Mycobacterium tuberculosis subsp. bovis bacillus Calmette-Guérin (BCG) carries BlaC and produces similar levels of BlaC as other M. tuberculosis strains (28). BCG was cultured as described previously (40) in 7H9 broth (Difco, Detroit, MI) supplemented with 0.5% glycerol, 10% oleic acid dextrose complex without catalase (OADC), and 0.05% Tween 80. Bacterial cultures were incubated at 37°C in the presence of 5% CO 2 until an optical density at 600 nm of 0.5. The bacterial culture was centrifuged and washed with 7H9 liquid medium without any supplements and resuspended in this same medium prior to use.
Purification of BlaC and other ␤-lactamases. Expression and purification of BlaC and TEM-1 were carried out as described previously (47,49). Expression and purification of AmpC, P99, KPC-2, and CTX-M-15 ␤-lactamases were carried out and provided by AstraZeneca (Waltham, MA). Activity of purified enzymes was validated and quantified by nitrocefin assays as described previously (59). Briefly, 0.2 nM of each enzyme in 2-(N-morpholino)ethanesulfonic acid (MES), pH 6.0, was incubated with 200 M nitrocefin at room temperature (25°C) to estimate active enzyme units. All activity assays were carried out in 100 l final volume. Activity of all enzymes used was determined the same day or within a few days of use. No loss of enzyme activity has been observed under standard storage conditions (47).
␤-Lactamase activity assays. Equal units, as determined by nitrocefin activity, of each ␤-lactamase were used to compare the specificity of CDG-3 with CDG-1 and CDG-OMe. Two ␤-lactamases, TEM-1 (class A ␤-lactamase) and AmpC (class B ␤-lactamase) were examined using various concentrations of enzyme up to 500 nM. All enzyme assays were conducted at room temperature (25°C). Briefly, 0.1 M fluorogenic substrate was incubated with different concentrations of ␤-lactamase for various periods of time in the dark. Enzyme and substrate dilutions were prepared in MES buffer (pH 6.0), and the assay was carried out in 96-well plates using an EnVision plate reader (PerkinElmer, Waltham, MA). Emission spectra upon excitation at 490 nm were collected and compared with and without BlaC. Standard nitrocefin assays (59) with various concentrations of BlaC were first carried out to standardize enzyme concentrations, and then dilutions were made and activity was determined in sputum, sputum with 200 mM MES buffer, pH 6.0, plus 2% dithiothreitol (DTT), designated TSS, and MES buffer using each of the fluorogenic substrates.
Detection of BlaC in sputum. Tuberculosis-negative human sputum was obtained from Robert Fader at Baylor Scott & White (Temple, TX). Sputum samples were stored at Ϫ80°C, thawed at room temperature (25°C), and mixed with an equal volume of TSS. ␤-Lactamases diluted in MES buffer were added (final concentrations of 100 to 500 nM) and incubated for 1 h. After 1 h of incubation, fluorogenic substrates in REF solution (49) were added to each sample (final concentration of 1 M). The subsequent change in fluorescence of the solution was monitored spectrophotometrically with excitation at 490 nm and emission at 515 nm. All of the assays, unless noted otherwise, were carried out in 96-well plates in triplicate.
Detection of bacteria in sputum. BCG was used as a control for the ability to detect M. tuberculosis in sputum as described previously (49). After estimating bacterial numbers by optical density at 600 nm, ϳ10 7 CFU of BCG were centrifuged, the supernatant removed and resuspended in the same medium. Dilutions were made, and ϳ100 CFU were added to pooled tuberculosis-negative human sputum. Samples were diluted 1:1 in TSS and incubated for 1 h at room temperature (25°C). Substrate concentrations varied from 0.01 M to 10 M in 200 mM MES, and samples were read immediately after adding the substrate and every 20 min for 3 h at room temperature in 24-well plates using a Mithras LB 940 plate reader (Berthold Technologies, Oakridge, Tennessee) at 490 nm excitation and 515 nm emission. Tuberculosis-negative sputum without BCG was included as a negative control in all experiments.

M. tuberculosis
BlaC is an unusual ␤-lactamase. The BlaC ␤-lactamase produced by M. tuberculosis is a biomarker for the presence of viable bacteria during infections in humans and animals (28,40,49,50). This enzyme is secreted through a twin-arginine translocation (Tat) pathway (60,61), resulting in BlaC being primarily cell associated and membrane localized (28). ␤-Lactamase activity in M. tuberculosis strains is almost exclusively due to BlaC, since mutations in the blaC gene lead to production of negligible ␤-lactamase activity by the resulting strains (28,61). Based on examination of microarray data in the Gene Expression Omnibus (GEO) at the National Center for Biotechnology Information (NCBI) and ArrayExpress at the European Bioinformatics Institute (EBI) databases, blaC expression is constitutive and not influenced by growth conditions. During infections, there are no significant differences between the first 24 h postinfection (62), under aerobic conditions (63), or during an oxidative stress response (64). Comparison of BlaC with tuberculosis complex members, those mycobacteria that cause tuberculosis, finds conservation throughout the protein ( Fig. 2A). BlaC is a class A ␤-lactamase since it carries the three motifs characteristic of this enzyme class (47,54). Class A ␤-lactamases from other bacteria have little similarity to BlaC outside of these motifs (Fig. 2B). Interestingly, the glycine near motif I and the three glycines near motif II are not present in other ␤-lactamases, and they map to the binding pocket in the crystal structure (47,48). These unusual glycines within the BlaC active site are what make the M. tuberculosis ␤-lactamase unique. BlaC is not found in other mycobacterial species or any bacterial species commonly found in the human oropharynx (65) (Fig. 2C). Overall, these observations suggest BlaC is specific to tuberculosis complex members, even within mycobacterial species.
CDG-3 is a highly specific and sensitive fluorogenic probe for M. tuberculosis. We designed three fluorogenic substrates that can sensitively detect BlaC, CDG-1, CDG-OMe, and CDG-3 (48)(49)(50). The specificity of these substrates is important because of the need for high specificity in patient samples where many other ␤-lactamases are likely present (48). Although specificity of CDG-OMe compared to TEM-1 (class A) ␤-lactamase has been examined (48), these substrates have not been directly compared for temporal kinetics to each other or with other ␤-lactamases. When we compared all three substrates directly, we found that the cephalosporin-based fluorogenic substrate CDG-1 is cleaved more readily by TEM-1 (class A) than by BlaC and AmpC (class C) ␤-lactamases (Fig. 3A). CDG-OMe, with a methoxy substitution at the 7-amino position, is cleaved more readily by BlaC than TEM-1 and AmpC (Fig. 3B). Selectivity of CDG-OMe for BlaC over TEM-1 is much better than that over AmpC (Table 1). CDG-3, with both a methoxy substitution at the 7-amino position and a cyclopropyl substitution at the 2 position, displays greater selectivity for BlaC than CDG-OMe (Fig. 3C). We found that CDG-3 displays high specificity for BlaC over other ␤-lactamases, even KPC-2, which is thought to be very versatile (66,67). Based on these data, we examined the use of CDG-3 to detect M. tuberculosis infection. Detection of BlaC enzyme in clinical material. The primary clinical material used for diagnosis of tuberculosis is sputum used for acid-fast staining and culture (1,22,52,68). We assessed the ability of the fluorogenic substrates to detect purified BlaC spiked into tuberculosis-negative human sputum. CDG-1, CDG-OMe, and CDG-3 allow sensitive detection of BlaC, and the signal correlates well with concentration of enzyme (Fig.  4A). Interestingly, human sputum interferes with detection of BlaC (Fig. 4B). We found that stabilization of pH with MES buffer and inclusion of DTT, designated TSS, allows quantitative detection of BlaC in sputum (Fig. 4C). In the case of sputum with TSS, CDG-1 and CDG-3 generated a greater signal from BlaC than CDG-OMe. We found that there was a very strong correlation (r 2 ϭ 0.99966; P ϭ 0.016) between BlaC concentration in sputum and fluorescence (Fig. 5). The correlation of fluorescence with the concentration of BlaC is similar in sputum and MES buffer (r 2 ϭ 0.99792; P ϭ 0.041). The reaction of BlaC with CDG-3 occurs rapidly in sputum, generating a nearly maximal fluorescent signal within the first 20 min (Fig. 5C). We found that, similar to cleavage of CDG-3 by BlaC in buffer alone (49), CDG-3 in sputum with TSS gives a 238-and 192-fold increase in fluorescence at 515 and 535 nm emission, respectively (Fig. 5D). We were interested in using 535 nm, even though similar data would be obtained in vitro at both wavelengths, since examination of clinical material by ourselves and other investigators suggests that collecting emission at longer wavelengths helps to reduce interference from hemoglobin that may sometimes be present in sputum (40,49,69). These data suggest that CDG-3 can be used to accurately detect M. tuberculosis BlaC in sputum, making it an excellent candidate for development of an REF diagnostic assay for tuberculosis.
REF assays can detect 100 CFU of M. tuberculosis in sputum. Culture can detect approximately 10 to 100 CFU of M. tuberculosis in sputum (70), making it important to optimize REF near this threshold. We have previously shown that tuberculosis-negative human sputum has large numbers of ␤-lactamase-producing bacteria (48,49), making it necessary to validate the performance of any detection strategy directly in human sputum rather than laboratory medium. We were particularly interested in using assays at ambient temperatures, since a diagnostic strategy that does not require temperature control would be advantageous for low resource settings where tuberculosis diagnosis is most needed (52). REF assays detect 100 CFU of BCG, which expresses similar levels of BlaC to those of other M. tuberculosis strains (28,40,49,50), spiked into human sputum within 20 min (Fig. 6). Interestingly, low concentrations of CDG-3 (0.01 M) display improved signal to noise ratios, allowing use of very little substrate for assays. We selected 1 M as the concentration for use in REFtb assays to ensure sufficient substrate is available in the complex environment present in clinical material where substrate may not always be readily accessible to the enzyme in solution. The sensitivity of these assays supports the feasibility of REF as a diagnostic assay for detection of tuberculosis infection using clinical material and suggests that its sensitivity may be similar to or better than that of culture.
Evaluation of REFtb for diagnosis of tuberculosis in patients. We obtained 160 sputum samples from suspected tuberculosis patients, coded and blinded to evaluate the performance of REFtb as a diagnostic test for tuberculosis. Sputum samples were obtained from a geographically diverse population, from 16 countries, including the United States. (Fig. 7A). When we used a cutoff value threshold for the highest frequency of correct predictions of tuberculosis diagnosis, the REFtb assay yielded an overall sensitivity of 89% for smear-and culture-positive samples and 88% for smearnegative and culture-positive samples (Table 2), with a specificity of 82%. We analyzed REFtb diagnosis against smear and culture combined and applied receiver operating characteristic (ROC) curve analyses. ROC for all samples gives an area under the curve of 76.27% (Fig. 7B), demonstrating significant (P ϭ 0.0002) correct predictive ability for REFtb diagnosis. The study population was 37.5% female, and no significant differences were observed for REFtb with respect to gender. REFtb performed similarly for all age  (Table 2). However, specificity was slightly higher for patients Ͻ35 years old and Ͼ65 years old than for patients 36 to 65 years old ( Table 2; P Ͻ 0.05). Clinical samples were assayed in 10 different batches of 7 to 20 samples each over 4 months. We found that there was variability in the signal obtained between batches for all samples in a batch. Thus, we explored the possibility that batch-specific analysis using the tuberculosis-negative sputum internal control in each batch for determining the cutoff threshold would be the best strategy for analysis of our data. When batch-specific analyses using the internal control from each batch are used, we obtain a sensitivity of 88.1% (95% CI ϭ 77.1 to 95.1) and a specificity of 86.1% (95% CI ϭ 77.8 to 92.2). The positive predictive value is 79% (P Ͻ 0.0001), and the negative predictive value is 93% (P Ͻ 0.0001). These data demonstrate that REFtb can diagnose tubercu- losis with high sensitivity, and batch-to-batch variability can be corrected through batch-specific analysis using tuberculosis-negative sputum controls.

DISCUSSION
We report evaluation of BlaC as a biomarker that is conserved within the tuberculosis complex for rapid diagnosis. The unique characteristics of the BlaC active site, constitutive expression, and conservation in the tuberculosis complex make it an excellent biomarker (48)(49)(50). The REFtb assay requires no processing and can be accomplished in 10 min. We show that the CDG-3 probe represents a substantial improvement over other probes in that it is stable and shows negligible cross-reactivity with other bacteria. Very small amounts of CDG-3 can be used in these assays, making REFtb a very inexpensive test, expected to cost less than a dollar per sample. Affordability and simplicity make REFtb a promising option for POC use that would allow more TB cases to be identified. Examination of 160 clinical specimens from suspected tuberculosis-infected patients yielded a high sensitivity and specificity of 88.1% and 86.1%, respectively, obtained from ROC analyses with high statistical significance (P ϭ 0.0002) and correct predictive value. These observations suggest REFtb is more sensitive than smear microscopy, which has a sensitivity of 20 to 80% (4). The negative predictive value of REFtb is 93% (P Ͻ 0.0001), suggesting that few cases are missed and emphasizing its potential for use as a rapid triage test. These values for specificity and sensitivity are directly in line with recommendations from the WHO for a biomarkerbased triage test to identify suspected TB patients (71). We are particularly excited by the observation that REFtb can detect the majority of smear-negative cases that would be missed in diagnosis prior to obtaining culture results, since about 17% of tuberculosis transmission is thought to occur from these patients (72). Implementation of REFtb a Clinical samples were divided into groups by age and sex to evaluate whether any of these differences impact the performance of the tuberculosis reporter enzyme fluorescent (REFtb) diagnostic assay. The total number of samples (n) in each group is shown in parentheses. b The percentage of female patients within the group. The number of samples in this group that were from female patients is shown in parentheses. at the POC has the potential to facilitate early diagnosis of tuberculosis, potentially reducing diagnostic costs by 34 to 43% in the developing world (73).
REFtb fills a unique biological niche in diagnostic assays since it is a catalytic enzyme-based assay. Since BlaC is secreted by the Tat pathway that requires ATP (60), REFtb is amenable to evaluating therapeutic outcomes and phenotypic drug susceptibility tests (DST). Existing diagnostic strategies and most others in development are based on different biological mechanisms (52), either nucleotides (GeneXpert), cell wall staining characteristics (smear microscopy), bacterial replication (culture), the host response (QuantiFeron-TB), or antibody-based recognition of molecules produced by M. tuberculosis. These differences in the biological mechanisms make it likely that REFtb will serve as a diagnostic strategy that is complementary to others. While data obtained in this initial evaluation of REFtb for diagnosis of tuberculosis in clinical samples are promising, it is still early in development. A larger number of clinical samples from diverse regions should be examined to confirm the REFtb sensitivity and specificity, though this study produced high significance values (P ϭ 0.0002). Optimally, batch variability should be eliminated to simplify analysis and prevent the need for inclusion of control samples. Batch variation is likely from preparation of the reagents, which currently are prepared fresh as liquids. Liquid reagents are not optimal for POC assays in limited resource settings. Stable lyophilized reagents are in development for REFtb that are expected to perform consistently from batch to batch, allow stable shipping at ambient temperatures, and eliminate solution preparation, facilitating use as a POC diagnostic assay.
Although the specificity of REFtb is very reasonable, particularly for a triage test in early development, some false-positives were obtained as defined by being negative by both smear and culture. There were no significant differences in the sensitivity between the age groups examined in this study, but there are small differences in specificity from the 36-to 65-year-old group that point toward a factor that can increase false positives. One possibility is that since REFtb is very sensitive, able to detect 1 to 10 bacteria in sputum (48)(49)(50), it is possible that some "false positives" are actually "true positives" that are missed by other tests and only identified by REFtb. Currently, we cannot differentiate true positives from true false positives, though follow up would allow us to differentiate them. With the sensitivity of smear microscopy at 20 to 80% (4) and most likely 50 to 60% (18, 74) and a single culture providing a sensitivity of ϳ80% (15,18), true positives can be missed by existing diagnostic strategies. We will conduct follow up in future studies that should allow us to differentiate true positives from false positives, but in the meantime, we are investigating other reasons.
Another possible explanation for false positives is that an unknown ␤-lactamase has a similar active site to BlaC or at least a similar ability to cleave CDG-3. We are investigating this hypothesis and, once a cross-reactive ␤-lactamase(s) is identified, incorporation of a specific inhibitor from the multitude of ␤-lactamase inhibitors available (75) and/or a specific antimicrobial that kills the cross-reacting bacterial species could possibly be used to eliminate these false positives and further improve REFtb specificity. Interestingly, the sensitivity of REFtb compared to both smear-positive and smear-negative tuberculosis samples is similar. This is most likely due to our choosing the most sensitive conditions possible for REFtb assays, where we read the samples after cleavage of the fluorogenic substrate plateaus, rather than while it is increasing linearly and remains quantitative. We previously observed this phenomenon with REF assays in vivo, where early time points allow quantitative determination of M. tuberculosis, but at later time points a similarly high signal is observed regardless of bacterial numbers present (28). Similarly, use of earlier time points in REFtb assays would likely allow quantitative measurement of bacterial numbers in sputum but would also most likely decrease sensitivity. Since we envision this test as a rapid triage test for tuberculosis, quantitation is less important than sensitivity, so we plan to retain the current assay conditions and focus on ensuring maximal sensitivity. Although false negatives are not common, we plan to investigate potential reasons that these occur to further improve the REFtb assay. One possible reason that samples that are positive by culture could become negative is improper storage and/or transport. This issue can be overcome by ensuring careful handling or by simply conducting the REFtb assays at the POC where bacterial viability is maintained. Since REFtb assays measure viable bacteria, we expect REFtb to perform even better on fresh samples than on samples that have been stored and shipped, even though the promising results from the current study were obtained using shipped samples. We envision REFtb as a triage test to be employed at the POC when a patient first visits a clinic. Management after identification may ultimately involve treatment or careful monitoring in parallel with additional testing, but with a negative predictive value of 93%, REFtb could be used to focus resources on those patients most likely to have tuberculosis.