Rapid Diagnosis of Smear-Negative Tuberculosis Using Immunology and Microbiology with Induced Sputum in HIV-Infected and Uninfected Individuals

Rationale and Objectives Blood-based studies have demonstrated the potential of immunological assays to detect tuberculosis. However lung fluid sampling may prove superior as it enables simultaneous microbiological detection of mycobacteria to be performed. Until now this has only been possible using the expensive and invasive technique of broncho-alveolar lavage. We sought to evaluate an immunoassay using non-invasive induced-sputum to diagnose active tuberculosis. Methods and Results Prospective cohort study of forty-two spontaneous sputum smear-negative or sputum non-producing adults under investigation for tuberculosis. CD4 lymphocytes specific to purified-protein-derivative of Mycobacterium tuberculosis actively synthesising interferon-gamma were measured by flow cytometry and final diagnosis compared to immunoassay using a cut-off of 0.5%. Sixteen subjects (38%) were HIV-infected (median CD4 count [range] = 332 cells/µl [103–748]). Thirty-eight (90%) were BCG-vaccinated. In 27 subjects diagnosed with active tuberculosis, the median [range] percentage of interferon-gamma synthetic CD4+ lymphocytes was 2.77% [0–23.93%] versus 0% [0–2.10%] in 15 negative for active infection (p<0.0001). Sensitivity and specificity of the immunoassay versus final diagnosis of active tuberculosis were 89% (24 of 27) and 80% (12 of 15) respectively. The 3 positive assays in the latter group occurred in subjects diagnosed with quiescent/latent tuberculosis. Assay performance was unaffected by HIV-status, BCG-vaccination or disease site. Combining this approach with traditional microbiological methods increased the diagnostic yield to 93% (25 of 27) alongside acid-fast bacilli smear and 96% (26 of 27) alongside tuberculosis culture. Conclusions These data demonstrate for the first time that a rapid immunological assay to diagnose active tuberculosis can be performed successfully in combination with microbiological methods on a single induced-sputum sample.


INTRODUCTION
Prompt detection of tuberculosis (TB) infection is vital for TB control [1]. The development of commercial interferon-gamma (IFN-c) release assays (IGRA) has focussed attention on the use of host immune reactivity as a marker of infection. Increasing data suggest that blood-based IGRA have significant advantages when investigating latent TB infection (LTBI) compared to the much older immunoassay the tuberculin skin test (TST) [2,3,4,5]. However, both the TST and IGRA are much less helpful in the diagnosis of active TB, especially when this is accompanied by immune suppression such as HIV [6,7,8].
Distinct TB antigen-specific responses are found in the lung [9,10]. Given that TB infection is transmitted typically by inhalation, it is likely that these responses, which are typically of a much higher frequency than found in blood, may be of greater importance than the immunological findings within peripheral blood. Clinical studies using bronchoscopy and broncho-alveolar lavage (BAL) to recover lung cells have confirmed that the pulmonary T-lymphocyte immune response can be used with excellent sensitivity to detect active TB, including non-pulmonary and HIV-related forms [11,12]. These early data have been confirmed in larger studies of spontaneous sputum-smear negative cases. [13,14].
Not only is a lung-based approach an exciting advance in TB immune-diagnostics; but also it allows direct smear and culture to be performed on the same primary sample. Given persisting high global rates of drug resistance and the emergence of XDR-TB, the facility to run such immunology and microbiology tests on a single sample is of great potential significance [15].
Obtaining lung fluid by bronchoscopy and BAL is both expensive and invasive [16]. Recently a number of studies have shown that non-invasive sputum induction (SI) using inhaled hypertonic saline for TB diagnosis has a similar microbiological yield to BAL-yet at a substantially reduced cost, and in a significantly more patient-friendly manner [17,18]. However, cell populations obtained by the two methods differ, which may limit assessment of TB antigen-specific responses when using SI [19].
In this study we have investigated for the first time whether a single sample obtained by simple SI methodology can be used to diagnose TB promptly using immunological assays (flow cytometry and ELISpot) in combination with traditional microbiological techniques.

METHODS Subjects
The study was approved by the Royal Free Hospital Ethics Committee (LREC 6236). Adults being investigated for TB who were either (1) sputum acid-fast bacilli (AFB) smear-negative, or (2) not producing sputum and had no positive AFB smears from nonpulmonary samples, were recruited from our urban Teaching Hospital between April 2005 and April 2006. Written informed consent was obtained. Subjects continued to be investigated by their physician independently of SI where appropriate. No subject had commenced anti-tuberculosis medication at the time of study.
SI was performed in a portable, negative-pressure isolation chamber (Elwyn Roberts, Shropshire, UK) using 3% saline delivered via an ultrasonic nebuliser (Sunrise Medical, Wollaston, UK) for 20 minutes. Peak expiratory flow rate (PEFR) was measured every 5 minutes and induction terminated if PEFR declined by .15%, or if chest discomfort were reported. Prior to induction subjects were asked to clear their mouth and nose of any residual saliva or secretions. After each five minute period and prior to any expectoration of sputum this process was repeated. Sputum was expectorated without assistance in to a sterile container.

Sample handling and laboratory investigation
Sputum was processed within 2 hours. An equivalent volume of 0.1% dithiothreitol solution (Sigma, UK) was added, the sample rolled for 20 minutes at room temperature and an aliquot (usually 50% of the total sample) removed for microbiological investigation.

Flow cytometry
Absolute leucocyte, lymphocyte and CD4+ T-cell numbers were determined by flow cytometry [20]. Detection of IFN-c synthesis following 16 hour (overnight) incubation was performed as previously described [11]. In brief, 5610 4 CD4 lymphocytes were incubated with tissue culture medium (TCM) alone {negative control}; TCM with PPD (Mycos Research, CO, USA) or TCM with PHA (Sigma) {positive control}. Brefeldin A (Sigma) was added after 2 hours. Harvested cells were fixed and permeabilised, and stained for expression of CD3, CD4 and IFN-c. Lymphocytes, identified by scatter, were gated according to CD3 expression to produce a histogram of IFN-c producing CD4+ and CD4-T cells. The IFN-c+ CD4+ lymphocyte frequency as a percentage of the total CD4+ T-lymphocyte population (%CD4+IFN-c+) was determined. The %CD4+IFN-c+ in the negative control was subtracted from that in the PPD-stimulated sample to give a percentage value of PPD-specific CD4+ IFN-c+ lymphocytes (%PPD-specific CD4+IFN-c+). No samples were excluded due to high negative control responses which had a median %CD4+IFN-c+ value of 1.04%. This did not differ significantly between subjects with and without active TB (1.04% versus 0.83% respectively). Using data from early experiments an adequate positive control response was taken as a CD3+ IFN-c+ response of .10% (median [range] positive control response was 49.3% [16.0-78.3]). All flow cytometry and evaluation of %PPD-specific CD4+IFN-c+ responses was performed by RAMB.

ELISpot
ELISpot assays were also performed on induced sputum samples as described previously [21]. Briefly 1610 5 leukocytes per well were cultured overnight with duplicates of TCM alone [negative control], PPD or PHA [positive control] in 96 well anti-IFN-c (Mabtech, Stockholm, Sweden) coated PVDF-backed plates (Millipore, Watford, UK). After overnight incubation IFN-c spot-forming cells (SFC) were detected according to the manufacturer's instructions (Mabtech). The mean of duplicate conditions was calculated. Results are presented as background subtracted (delta) spot-forming units (DSFU) per/10 6 cells. A negative response to the positive control, PHA, was used to determine assay failure, in which case the result was discounted from the data set. Since none of the PHA treated cells produced a negative result, no assays were excluded. All ELISpot assays and determination of these results were performed by GADH.

Microbiology
Microscopy acid-fast bacilli (AFB) smears were prepared and stained with auramine-O; and positive smears were confirmed using a Ziehl-Neelsen stain. All positive cultures for Mycobacterium tuberculosis (M.tb) were confirmed by the Mycobacterial Reference Laboratory, London, UK.

Diagnostic definition
Clinical diagnostic decisions were made without reference to immunological data. A final diagnosis of ''active TB'' was accepted when (1) M.tb was cultured or (2) there was clear evidence for a clinical diagnosis based on radiology or histology with an appropriate response to treatment. A final diagnosis of ''quiescent or latent TB'' was accepted if there were a clearly recorded history of exposure to a smear positive index case with supporting tuberculin skin test (TST) or radiological results but the clinical decision had been made not to commence full anti-tuberculosis therapy. A final diagnosis of ''not TB'' was accepted if an alternative diagnosis were reached, symptoms resolved rapidly and anti-tuberculosis therapy was not commenced. All subjects without culture-confirmed TB have been followed up for at least 12 months for changes in their diagnostic categorisation.

Statistical assessment
Comparisons between groups were made using the Wilcoxon test. The data were analysed using SAS version 8.2 (SAS Institute Inc, Cary, NC).  Table 1.
Three of 42 subjects received a final diagnosis of quiescent or latent TB. A final diagnosis of ''not TB'' was made in 12 subjects, all of whom had been previously BCG-vaccinated and 7 of 12 (58%) were HIV-infected. These data are shown in Table 2.

Induced-sputum as a diagnostic tool-Immunoassay alone and in combination with AFB smear or TB culture
From ROC curve analysis (data not shown), a post-hoc %PPDspecific CD4+IFN-c+ value of $0.5% was selected to define a positive assay. Using this, the assay had a sensitivity of 89%  Table 2.

DISCUSSION
This is the first study to demonstrate that a single induced-sputum sample can be used for combined immunological and microbiological TB diagnosis. In 42 individuals (38% of whom were HIVinfected) who were either not spontaneously producing sputum or in whom this was AFB smear-negative, our immunoassay had a sensitivity of 89% (24 of 27 positive) and specificity of 80% compared to a final clinical diagnosis of active TB. Although somewhat smaller numbers, these results compare very well with those achieved in blood in HIV-uninfected subjects [7], and ours reported previously using BAL in a similar population [14].
Combining our lung-orientated immunological approach with traditional microbiology produced prompt identification (inducedsputum AFB smear or immunoassay-positive) in 93% of cases; and overall identification (induced-sputum TB culture or immunoassay positive) in 96% of those with active TB at any body site. It should be noted that of the 15 subjects with active pulmonary TB, 4 were AFB smear-positive and 13 were immunoassay-positive on a single induced sputum sample. Thus the latter greatly increased the chance of prompt diagnosis of active, infectious pulmonary TB. The induced sputum immunoassay was also unaffected by HIV status, TB disease site, and BCG vaccination. The similar results obtained using the different methodologies of flow cytometry and ELISpot also demonstrate that the potential of this sputum-based method is not limited to one technique.
Confounding positive responses from BCG-vaccinated individuals without TB have been observed using both the tuberculin skin test (TST) and PPD-based blood assays [22]. This has led to the use of smaller antigens encoded in the region of difference 1 (RD1) portion of the genome of M.tb, which is absent in BCG-strains [23]. However, in our lung-based assay, despite all 12 subjects with a clinical diagnosis of ''not TB'' having received BCG, none had evidence of a detectable pulmonary PPD response. We believe this striking finding reflects the organ-specific pathway taken by vaccination-generated BCG-specific memory lymphocytes ie. trafficking through blood and homing to the skin whilst avoiding the lung [24,25]. Apart from the insight this offers for future TB vaccination strategies, we feel that these data support the use of PPD, which as a rich antigen mixture, generates higher frequency responses than RD-1 encoded antigens and therefore is likely to be beneficial in the investigation of a condition associated with immune suppression and CD4 lymphopenia in HIV-infected and HIV-uninfected individuals [26,27,14].
The sub-optimal specificity observed represents positive assays occurring in 3 subjects with a final diagnosis of quiescent or latent TB. Interestingly these individuals had co-existent conditions which traditionally make TB diagnosis difficult (2 HIV-infection, and 1 terminal malignancy) and reduce the sensitivity of the TST. Given this, it could be argued that a positive result is not such a bad thing as it alerts the clinician not to discount TB. That positive results are found in subjects with both quiescent or latent TB and those with active TB confounds all current immune-based assays using interferon-c expression as the primary readout [28,29]. However we have previously observed that some of the individuals with these apparent ''false-positive'' results may develop clearly active TB during follow up [14].
When seeking to delineate between different stages of TB we believe that flow-cytometry adds more value than other immunebased tests, as it can define not only the numbers of cells producing a given cytokine but also the cell phenotype responsible in various states of TB infection. Early studies assessing the expression of both interferon-c and surface markers of differentiation and memory status such as CD27 have hinted at the promise of such a multi-parameter flow-cytometry based approach [30].
The focus of this study has been the assessment of our inducedsputum assay in diagnosing active TB. The methodology is applicable also to studies of TB pathogenesis. To date, lung-based immunological studies have relied upon bronchoscopy. Although this usefully obtains cell-rich samples, its invasive nature makes it an unpleasant test. Sputum induction offers the possibility of longitudinal assessment of the pulmonary immune response during treatment with good subject acceptability and at greatly reduced cost. We believe that this approach opens up an important area of research which might yield markers of treatment response and even TB cure (31).  These data reveal for the first time how a lung-orientated approach combining a novel immunoassay with traditional microbiology can rapidly diagnose active TB in individuals with smear-negative disease using a single induced-sputum sample. Although a relatively small study performed in a research setting, we believe that our data warrant further work assessing the usefulness of induced-sputum as an alternative to both blood and BAL samples when investigating possible TB.