Sputum Microscopy With Fluorescein Diacetate Predicts Tuberculosis Infectiousness

Summary Fluorescein diacetate microscopy was used to assess Mycobacterium tuberculosis metabolic activity in pretreatment sputa. Lower results predicted greater infectiousness in transmitting tuberculosis to household contacts. Fluorescein diacetate microscopy–negative bacteria may be better adapted to airborne tuberculosis transmission and consequently more contagious.

Patients with pulmonary tuberculosis (TB) expectorate sputum that contains heterogeneous subpopulations of Mycobacterium tuberculosis (Mtb) [1], which has unknown implications for assessing patient infectiousness [2]. Currently, predictions of TB infectiousness are largely based on sputum acid-fast smear microscopy results that are used to guide decisions about infection control measures and to prioritize which patients' contacts should receive screening, chemoprophylaxis, and surveillance [3,4]. However, acid-fast microscopy stains Mtb bacilli whether they are metabolically active, inactive, or dead [5]. This is important because in vivo and in vitro experiments have suggested that Mtb bacteria that have a slowly metabolizing phenotype induced by growth under stress-inducing conditions are most infectious [6][7][8][9].
Fluorescein diacetate (FDA) is used in viability assays because it only visualizes cells that produce nonspecific esterase enzymes and have intact cell membranes to retain them, because these enzymes must be present to hydrolyze FDA to its fluorescent form [10]. FDA has been used with sputum microscopy to rapidly and inexpensively predict TB culture positivity [11,12], and we have found that this can predict the amount of culturable TB in sputum [13] and assess whether early TB treatment is inducing an appropriate treatment response in patients [14,15].
For the present research, we hypothesized that microscopy with FDA of pretreatment sputum would predict TB infectiousness, as suggested in an editorial [16] commenting on our previous findings. To test this hypothesis, we worked with a group of patients with pulmonary TB who had FDA microscopy performed as previously described [14]. To assess whether FDA microscopy results predicted their risk of transmitting TB, we followed up their contacts and screened those contacts for TB. professionals in collaboration with the local ministry of health, as described elsewhere [14]. Patients received empiric first-line TB therapy with clinic-based direct observation of every dose provided by the national TB program, which follows World Health Organization guidelines [17]. Tuberculosis-related care for patients and their close contacts including diagnostic testing, treatment, and chemoprophylaxis is provided free from direct charges in Peru and was not delayed or modified by study participation.

Study Participants
Patients were unselected consecutive adults diagnosed with TB that was strongly sputum microscopy-positive (ie, Ziehl-Neelsen grade "+++" or "++, " excluding patients whose samples were "+" or negative [18]). Recruitment took place in shantytowns in Callao, Peru, where the average TB notification rate in these participating clinics during the study period was 194 cases per 100 000 people per year. According to local practice, we estimated that the background case notification rate in these communities was 20% higher (ie, 233 cases/100 000 people/year) because some cases were treated by other providers [17,19,20].
Patients' contacts were invited to participate if they spent ≥6 hours/week in the patient's household in the 2 weeks prior to the patient's diagnosis [21]. Patients and contacts were interviewed to record symptom and demographic data; weight and height were measured; and, for adults, body mass index (BMI) was calculated. Socioeconomic status was assessed by a composite household index in arbitrary units derived by principal components analysis from 13 variables characterizing education, housing conditions, basic services and assets, as described previously [21,22]. Radiographs were not routinely performed so data were not available [23]. Patient characteristics are shown in Table 1.

Contact TB Chemoprophylaxis
Isoniazid chemoprophylaxis was provided for 6 months free of charge by the national TB program to contacts aged <15 years, once the national TB program had excluded active TB disease, regardless of any tuberculin skin test (TST) results [23]. Older contacts, especially those aged 15-19 years, were also eligible to receive chemoprophylaxis at the discretion of the national TB program clinician [23].

Contact Tuberculin Skin Testing
For the current research study, when the patient with newly diagnosed TB was commencing therapy, contacts aged ≥15 years in their household were asked at the time of patient recruitment to undergo screening for asymptomatic latent TB with a TST, as described [24,25]. However, to encourage participation, our study did not require a TST.

Contact TB Prevalence Surveys
At the time of recruitment during 2006-2007, and again during follow-up visits approximately 3 and 6 years later until 2013, participants were asked whether they had TB diagnosed ( Figure 1). Self-reported TB episodes were confirmed by checking national TB program records. At these recruitment and follow-up visits, contacts were also asked whether they had symptoms suggestive of TB disease. Contacts with cough, fever, night sweats, or weight loss were requested to provide a sputum sample that we tested with acid-fast auramine microscopy [26] and the microscopic-observation drug-susceptibility (MODS) liquid culture technique [27,28].

Sample Processing
Prior to treatment initiation, a pooled single sputum sample was collected overnight for approximately 12 hours at room temperature, and transported at 4°C to our laboratory the following morning. Samples were decontaminated, centrifuged as described previously [14,26], and immediately smeared onto microscope slides. The pellet suspension was also inoculated for MODS quantitative culture including serial dilutions of each sample, acid-fast auramine microscopy, and FDA microscopy, all performed in triplicate, as described previously [14,[27][28][29][30]. All samples were also tested for isoniazid and rifampicin resistance, as described previously [14]. Microscopy techniques together took approximately 1 hour.

Bacterial Quantification
Visible stained bacteria were counted in 100 consecutive microscopy fields at high power ×1000 magnification with an oil immersion lens. These microscopy counts of bacteria and the numbers of colony-forming units (CFU) per culture well were transformed to the concentration per microliter of sample [14,29]. All laboratory tests were performed blinded to clinical status and to other test results [14].

Statistical Analysis
Statistical analysis was done with Stata software version 12, as described previously [14]. Bacterial and CFU counts were exponentially distributed, so were transformed to their base-10 logarithm. Tests were 2-tailed and were performed with a 95% confidence level. Data with normal distributions were summarized by mean and standard deviation (SD) and nonparametric data by median and interquartile range (IQR). Time-to-event Cox regression analysis was used to analyze the hazard ratio (HR) of TB diagnosis in contacts adjusted for household clustering, censored at the time of last follow-up. For the time-to-event Cox regression analysis, secondary TB was a rare outcome and independent variables did not have Gaussian distributions, so all data were analyzed as dichotomous variables above or below the median value for contacts, labeled "high" or "low, " respectively. All univariable analyses, including univariable Cox regression analyses, were performed with nonimputed data. To prevent missing data in >1 variable from reducing the statistical power of multivariable analysis, the primary multivariable analyses were performed with imputed data. Data were imputed by extrapolating results from corresponding laboratory results, or by using the median response from the participants' questionnaire data. Two multivariable sensitivity analyses were also done: (1) excluding imputed data and (2) censoring follow-up after 5 years.

Laboratory Characterization
All 35 pretreatment sputum samples from patients had positive microscopy results for acid-fast auramine and FDA, and had confirmed Mtb in culture [14]. Quantitative culture results were obtained for 96% (33/35) of samples because 2 failed due to fungal overgrowth. Quantitative culture CFU results were predicted by both FDA microscopy (coefficient = 0.66; R 2 = 0.46; P < .001) and by auramine microscopy (coefficient = 1.1; R 2 = 0.47; P < .001; Table 1). Time to positive culture (TTP) in days was significantly associated with quantitative culture results and with FDA microscopy results (both P ≤ .01; Table 1). Drug susceptibility testing revealed that 11% (4/35) of patients had multidrug-resistant (MDR) TB. Two patients had isoniazid-monoresistant TB that we classified for statistical analysis as non-MDR-TB.

Patient Characteristics
The patient characteristics and their associations with the results of sputum laboratory testing are shown in Table 1. Constitutional symptoms (night sweats and/or fevers) were associated with high results of: auramine microscopy, FDA microscopy, and culture CFU (all P ≤ .04; Table 1). No patients had evidence of extrapulmonary or disseminated TB.

TB Infection in Contacts
At the time of patient recruitment, 209 of the patients' contacts were also recruited ( Figure 1). Contact characteristics are summarized in Table 2. These contacts had a median age of 20 (IQR, 10-32) years, 50% (105/209) were male, and the mean BMI was 25 (SD, 5.2) kg/m 2 . Only 40% (84/209) of the contacts underwent a TST. This was because 38% (80/209) of the contacts were ineligible because they were aged <15 years and 35% (45/129) of the remaining contacts declined. Seventy percent (59/84) of TST results were positive. Positivity was not associated with any laboratory or clinical characteristics (data not shown), but confidence intervals (CIs) were wide because so few participants had a TST.

TB Disease in Contacts-Univariable Analysis
TB disease was diagnosed in 6.2% (13/209) of contacts during follow-up that lasted a median of 6.3 (IQR, 6.2-6.5) years. This was significantly higher than the background community rate (HR, 4.0 [95% CI, 1.9-8.4]; P < .001, Figure 3A). All the variables shown in Table 1 and the contact characteristics were tested for associations with secondary TB disease in contacts using univariable Cox regression analysis. FDA microscopy-positive concentrations of bacteria per microliter were analyzed around the median value as high (median, 386 [IQR, 163-601]) or low (median, 22 [IQR, ). All variables with biologically plausible or statistical evidence (P < .1) of associations with secondary TB in contacts are shown in Table 2. Contacts of patients with low FDA microscopy results were 4-times more likely to have TB disease (unadjusted HR, 3.8 [95% CI, 1.1-13]; P = .03; Table 3 and Figure 3B). Specifically, 2.7% (95% CI, .57%-7.8% [3/110]) of contacts exposed to patients with high FDA microscopy results developed TB disease during follow-up, which did not significantly differ from the background community rate (P = .4; Figure 3). In contrast, 10% (95% CI, 5.0%-18% [10/99]) of contacts exposed to patients with the low FDA microscopy results developed TB disease during follow-up ( Figure 3B).
TB disease was more likely in contacts of patients who reported having a productive cough and in households with poorer than median socioeconomic score (both P ≤ .03; Table 2). Chemoprophylaxis was taken by 9.1% (19/209) of contacts, all aged <15 years, none of whom developed TB disease. Constitutional symptoms (fever and/or night sweats) in patients was associated with FDA microscopy results, and consequent impact on diagnostic delay may influence TB transmission [31,32], but including either or both variables in all multivariable regression analyses did not affect the significant association between lower than median FDA microscopy results and increased secondary TB. Tuberculosis disease in contacts was not associated with high vs low auramine microscopy results (P = .7; Table 2), but all patient samples were strongly auramine microscopy positive. Although in univariate analysis, quantitative culture CFU did not predict TB disease (P = .2;  Results were calculated by multivariable Cox regression analysis adjusting the association of FDA microscopy with the variables that in univariable analysis significantly predicted secondary tuberculosis (TB) disease in contacts (shown in Table 2). High vs low auramine microscopy results were not associated in univariable analysis with secondary TB (P = .6) and, if despite this auramine microscopy results were included in the adjusted analyses, then the pattern of significance was unchanged. The presence vs absence of multidrug-resistant (MDR) TB was not associated in univariable analysis with secondary TB (P = .3) and, if despite this MDR-TB was included in the adjusted analyses, then the pattern of significance was unchanged. The primary multivariable analysis included 5 imputed data for quantitative culture and productive cough (see Results section). Sensitivity analysis 1 excluded the imputed data. Sensitivity analysis 2 censored all follow-up at 5 years, causing 1 late case of secondary TB to be excluded.

TB Disease in Contacts-Adjusted Multivariable Regression Analyses
For the primary multivariable Cox regression analysis (Table 3), 5 missing data points were imputed. Two patients had missing CFU data because of fungal overgrowth, and high/low CFU categorization was imputed from days to culture positivity and auramine microscopy results. Three patients did not report whether they had productive cough, and this was imputed as the median response. The association between low FDA microscopy results predicting increased risk of TB disease in patients' contacts main-  Table 3).
Sensitivity analyses that excluded all imputed data or had follow-up censored after 5 years (that excluded the last episode of secondary TB in contacts that occurred 5.7 years after index case diagnosis and could have been unrelated to that exposure) had almost identical results to the analyses above (Table 3).
Post hoc exploratory analysis of TTP data considered patient cultures as faster or slower than the median value of 7 days.

DISCUSSION
In strongly microscopy-positive sputum samples collected before TB treatment, the proportions of FDA-stained and culturable Mtb bacteria varied considerably between patients. These findings are clinically significant because patients with low FDA microscopy results were most infectious, as demonstrated by the finding that their household contacts were considerably more likely to develop TB disease, including when adjusted for confounding variables and in sensitivity analyses.
Sputum from patients with pulmonary TB is known to contain heterogeneous subpopulations of Mtb bacteria [1,33,34]. Culture techniques used for diagnosing TB do not generally identify these individual subpopulations because results are determined by the overall quantity of bacteria as well as the fastest-growing subpopulation. A subpopulation of particular interest is persister cells, a term used in the context of chemotherapy to describe Mtb that persists in sputum in a slowly metabolizing, nonreplicating state [35]. In pretreatment sputum, most Mtb bacteria are believed to be in a persister-like state, as they cannot be cultured using conventional techniques and contain lipid bodies, both features of the phenotypic variant of Mtb bacteria that are slow growing or nonreplicating [1,33,36]. Furthermore, Mtb in persister-like states has been  demonstrated not to fluoresce with FDA, unless resuscitation-promoting factors are added [37]. Nonfluorescence with FDA could be explained by loss of cell membrane integrity, changes in the cell wall effecting FDA penetration, or low esterase activity when in a slow metabolizing state [35,38]. We found that patients were approximately 4-times more likely to cause TB disease in their household contacts if they had low concentrations of FDA microscopy-positive Mtb in their sputum. Mtb persister-like states develop in response to stressful environments such as nutrient deficiency or hypoxia [39], and persister-like phenotypes of Mtb are thought to be better adapted to surviving harsh environments such as during airborne transmission [1]. This increased infectiousness of Mtb in persister-like states has been demonstrated by growing Mtb under hypoxic or nutrient stress. Compared with Mtb grown under conventional aerobic conditions, in vivo stress-induced Mtb caused 10 times greater airborne infectiousness to guinea pigs and an altered immune response that increased pathogenicity to mice, and in vitro this increased invasion of epithelial cells and alveolar macrophages [6][7][8][9]. We hypothesize that in our study, which only included strongly sputum microscopy-positive TB, patients with low concentrations of FDA microscopy-positive Mtb in their sputum had high concentrations of persister-like mycobacteria that were better adapted to airborne transmission, so were more infectious. This hypothesis warrants specific research, in addition to confirmation of the association with infectiousness in larger studies and with other patient groups.
We found considerable variability in the proportion of FDAstained Mtb between different patients' pretreatment sputa. Most Mtb in sputum did not fluoresce with FDA microscopy, and could indicate dead bacteria. However, the most infectious patients in this study were those with low numbers of FDA-staining bacilli but high colony counts in quantitative culture. This indicates that a large proportion of FDA microscopy-negative Mtb were not dead but rather had a phenotype associated with greater infectiousness. It is possible that the use of resuscitation factors [37,40] or more prolonged culture [33] may cause higher concentrations of Mtb to be identified in both FDA microscopy and quantitative culture. Processing sputum without decontamination may also have this effect, but would increase fungal overgrowth of cultures, which confounds CFU quantification [29]. As expected, patients with more constitutional symptoms had significantly higher sputum Mtb concentrations, probably due to having more extensive pulmonary TB disease [41].
We found that the risk of secondary TB disease was higher for contacts of patients with high quantitative culture results, whether indicated by high CFU (in adjusted analysis) or rapid culture TTP (in univariable analysis), both indicators of mycobacterial load. Generally, smear positivity is also associated with greater infectiousness, but a meta-analysis of risk factors for TB transmission required thousands of participants to demonstrate that patients with concentrations of Mtb in their sputum sufficient for positive acid-fast microscopy had a 1.4 times (95% CI, 1.2-1.6) increased risk of causing TB disease in their contacts compared with acid-fast microscopy-negative patients [42]. This finding does not contradict the results of our study, which included only strongly microscopy-positive patients. In contrast to the 1.4 times greater TB infectiousness in acid-fast microscopy-positive vs acid-fast microscopy-negative patients in this meta-analysis, our results suggest that FDA microscopy results had a considerably larger effect that was independent of the results of microscopy, CFU, TTP, or other potentially confounding variables. In fact, our study demonstrates that contacts of patients who were acid-fast microscopy positive, but had high concentrations of fluorescing bacteria in FDA microscopy, had no higher risk than the background rate in their community.
Contacts of patients with pulmonary TB disease are at high risk of developing TB disease and are often provided with chemoprophylaxis. This policy is consistent with the apparent protective effect of chemoprophylaxis in our study. However, in most high-TB-prevalence settings, resources are insufficient for comprehensive contact tracing within and outside the home [43]. Provision of chemoprophylaxis to all household contacts may be appropriate in some settings. However, a low-cost tool, such as FDA microscopy, may identify the most infectious individuals and allow better prioritized contact tracing for TB disease screening and prevention than the current use of conventional sputum acid-fast microscopy results.
In addition to studying infectiousness to cause TB disease in contacts, we also aimed to assess infectiousness to cause asymptomatic latent TB in contacts. However, most of the contacts in our study were too young to be eligible for, or declined, a TST. This low uptake of TSTs reduced the statistical power so much that we were unable to meaningfully test for associations with asymptomatic latent TB. Furthermore, concurrently we discovered that the majority of healthy adults in these communities were already TST positive, even without living with a known TB patient [22,44]. Thus, living with a patient newly diagnosed with pulmonary TB increased the risk of TST positivity by <50% [22,44]. Therefore, to adequately test the hypothesis that FDAnegative Mtb in sputum also predicts infectiousness for causing asymptomatic latent TB would require much larger future studies, ideally using the more specific interferon-γ release assays in a setting with a low background rate of TB infection [44][45][46].
The main limitation of this research is that FDA microscopy was only assessed in Mtb strongly microscopy-positive samples. Therefore, in future research, FDA microscopy should be assessed in samples with weakly positive and negative acid-fast microscopy results, which may be facilitated by concentrating Mtb in sputum by filtration [47] or centrifugation [29]. To reduce the risk of false-positive FDA microscopy results due to other viable bacteria within the sample, we decontaminated all sputum samples, which is known to kill some Mtb [48]. Whether decontamination selectively affected the viable subpopulation that was not stained by FDA is unknown and warrants future research. However, this would be expected to affect all samples equally because the same protocol was followed for all the specimens processed, and should also similarly affect the quantitative culture results. Therefore, decontamination could not explain the contrasting results demonstrated between FDA microscopy and quantitative culture predicting infectiousness. Another limitation is that the Mtb TB strains were not available to use molecular fingerprinting to confirm whether any of the secondary TB cases in contacts were caused by TB patients other than the index case recruited to our study [49,50]. However, our conclusions were supported by a sensitivity analysis excluding the last episode of secondary TB in case that late case was unrelated to the index patient (and another sensitivity analysis excluding 5 imputed data points).
In conclusion, the inexpensive, rapid, low-technology technique of FDA microscopy revealed that only a small and variable subpopulation of Mtb in sputum was FDA microscopy positive. This study provides evidence that among strongly sputum microscopy-positive patients, those with low concentrations of FDA microscopy-positive Mtb were most infectious to cause TB disease in their contacts. This supports the hypothesis that a subpopulation of Mtb that do not stain with FDA, such as persister-like bacteria, are most infectious to transmit TB disease. Thus, FDA microscopy identified a high-risk group of TB patients whose contacts particularly need interventions to prevent and screen for secondary TB disease.

Notes
Author contributions. All authors participated sufficiently in the work, the data analysis, and/or the preparation of the manuscript, and all have reviewed and approved the manuscript as submitted to take public responsibility for it. Conception, hypotheses delineation, and study design: