SecA1 PCR on Sputum or Oral Wash for the Diagnosis of Pulmonary Tuberculosis

Abstract

Background

Nucleic acid amplification tests are sensitive and specific for identifying Mycobacterium tuberculosis in sputum-smear-positive populations, but less sensitive in sputum-smear-negative populations. Few studies have assessed their performance in patients with HIV, and no studies have assessed their performance using oral-wash specimens, which may be easier to obtain than sputum.

Methods

We performed a prospective study of 127 adults from two populations undergoing evaluation for respiratory complaints at Mulago Hospital in Kampala, Uganda. We tested sputum and simultaneously collected oral-wash specimens for Mycobacterium tuberculosis DNA by polymerase chain reaction(PCR) amplification at a novel locus, the secA1 gene. A positive sputum mycobacterial culture defined cases of tuberculosis(TB); we calculated sensitivity and specificity of the PCR assay on sputum or oral wash in reference to this gold standard.

Results

TB(75/127, 59%) and HIV(58/126, 46%) were both common in the study population. Sputum PCR was highly sensitive(99%, 95% confidence interval (CI) 93% to 100%) and specific(88%, 95% CI 77% to 96%) for pulmonary TB, and performed well in patients with HIV and in those with negative sputum smears. Oral-wash PCR was less sensitive(73%, 95% CI 62% to 83%) but also detected a substantial proportion of TB cases.

Conclusions

PCR targeting the secA1 gene was highly sensitive and specific for identifying M. tuberculosis in sputum, independent of smear or HIV status. Oral washes showed promise as an easily obtained respiratory specimen for TB diagnosis. SecA1 PCR on sputum could be a rapid, effective diagnostic tool at tuberculosis referral centers.

Introduction

Worldwide nine million new cases of active tuberculosis(TB) occur annually. In most high-burden countries, national reference laboratories rely on microscopy for diagnosis of TB. In practice, this labor-intensive procedure is insensitive, even when specimens are treated with mucolytic agents, concentrated by centrifugation, and examined using fluorochrome stains. Furthermore, microscopy is subjective and time-consuming, and guidelines recommend microscopists allocate at least five minutes to each slide and review no more than 25 slides per day.[1] Culture provides a definitive diagnosis of TB, but only after weeks of incubation, followed by identification using biochemical or molecular methods. Every step requires technical expertise for optimal laboratory performance.[2] Thus, an automated, molecular system for detecting M. tuberculosis(M.TB) would be especially useful for health care facilities in high-burden countries, eliminating the subjective interpretation of microscopy and inherent delays associated with culture.

Direct amplification of M.TB nucleic acids in sputum has been proposed as a means for improving the sensitivity of diagnostic testing for pulmonary TB.[3] Several commercial assays, such as the amplified M. tuberculosis Direct Test(Gen-Probe, San Diego, California), the Amplicor M. tuberculosis polymerase chain reaction(PCR) assay(Roche Molecular Systems, Branchburg, New Jersey), and the BD ProbeTec ET Direct Tuberculosis assay(Beckton Dickinson Diagnostics, Sparks, Maryland) are available for direct detection of M.TB in clinical specimens. While some studies from low-income countries have reported that sputum nucleic acid amplification is accurate and cost effective[4, 5], its performance for TB diagnosis has proved disappointing[6–8], with a pooled sensitivity of 72%.[9] Sensitivity is especially low in patients with negative microscopy results.[10] Although smear-negative TB cases in HIV-infected patients account for about 90% of undiagnosed TB in parts of sub-Saharan Africa[11], little has been published about how nucleic-acid-amplification assays for TB perform in patients with HIV.[12]

We previously reported that a PCR assay targeting the secA1 gene identifies the majority of mycobacterial species.[13] In the current study, we prospectively evaluated the performance of a M.TB-specific secA1 PCR assay in sputum for diagnosis of pulmonary TB in a Ugandan population with a high prevalence of HIV and TB. A parallel aim was to evaluate whether PCR on a specimen more easily obtained than sputum, oral wash with sterile saline, would provide satisfactory sensitivity and specificity for pulmonary TB, as previously demonstrated for Pneumocystis pneumonia.[14]

Methods

Participants

We prospectively enrolled outpatient and inpatient TB suspects at Mulago Hospital in Kampala, Uganda. Between September 2006 and May 2007, a parent study of voluntary counseling and testing recruited every third outpatient presenting for initial evaluation at the Uganda National Tuberculosis and Leprosy Programme(NTLP) clinic.[15] We invited the first two outpatients enrolled each day to join our study.

At a separate location on the medical wards, a research assistant recruited all inpatients ≥18 years of age admitted between March and May, 2007, with chronic(≥3 weeks but <6 months) cough or dyspnea, excluding those taking TB treatment. A second, blinded research assistant randomly selected two inpatients from eligible individuals for enrollment each day. Since the inpatient study began after the outpatient study, the fixed sample size for the combined study limited inpatient enrollment.

The Joint Clinical Research Centre Institutional Review Board approved the outpatient protocol. The Makerere University Research Ethics Committee and the Mulago Hospital Institutional Review Board approved the inpatient protocol. The Committee on Human Research at the University of California, San Francisco, and the Uganda National Council for Science and Technology approved both protocols.

Patient Data and Specimen Collection

After written informed consent, participants provided demographic and clinical information in response to a standardized questionnaire, and were offered HIV testing and counseling according to a standard protocol.[15] CD4+-T-lymphocyte counts were measured at enrollment in HIV-infected inpatients, but not in outpatients.

Each participant provided expectorated sputum at enrollment. Samples taken from outpatients were processed using 10% dithiothreitol(Prolab Diagnostics, Richmond Hill, Ontario). Specimens from inpatients underwent processing at the National Tuberculosis and Leprosy Program Reference Laboratory(NTRL) using 1% N-acetyl-L-cysteine(NALC), 2% sodium hydroxide(NaOH), and 2% sodium citrate.[16] For collection of oral wash, subjects were instructed to cough vigorously five times, and then gargle 10 mL of sterile normal saline for 60 seconds. Oral washes were processed using 1% dithiothreitol. Both sputum and oral-wash specimens were concentrated by centrifugation at 3000 × g, and the resulting pellet was frozen at −20° C within eight hours of collection. Specimens were stored for up to seven months(median 110 days, range 7 – 195 days) and sent frozen to the National Institutes of Health(NIH) Clinical Center. Upon arrival, acid-fast smear and mycobacterial culture were performed on sputum, and PCR was performed on sputum and oral wash.

Specimen Testing

M.TB PCR

All sputum and oral-wash specimens were tested with a PCR assay targeting the M. TB gene coding for the protein secA1(locus tag Rv3240c, GeneID 888860), a component of the major pathway of protein secretion across the cytoplasmic membrane.[13] Specimens were processed by ultrasonication for 15 minutes in a sodium-dodecyl-sulfate-Tris-HCl buffer with zirconia/silica beads(BioSpec Products, Bartlesville, Oklahoma) and then placed in NucliSens-lysis buffer(Biomerieux, Boxtel, The Netherlands) to rupture the mycobacterial cell wall and release DNA. DNA was purified using a NucliSens nucleic-acid-extraction kit. Real-time PCR on a LightCycler(Roche, Basel, Switzerland) was performed using M13-tailed primers to amplify a 490-nucleotide region of the secA1 gene found in all members of the Mycobacterium genus.[13] Two PCR reactions for each sample were performed in separate capillary tubes using different sets of FRET probes but the same primers. One set was specific for the M.TB complex; another set was specific for the Mycobacterium genus. A third set specific for a plasmid containing an internal-control sequence was run for each sample to detect PCR inhibition. If a specimen showed inhibition in the third tube without positive amplification of M.TB-complex, all three capillaries were re-analyzed using a 1:10 dilution of the specimen DNA. For each PCR run, a negative control(using the same primers and probes without template DNA) was run to detect contamination. Additional procedures to avoid DNA carry-over included unidirectional workflow, with separate rooms for extraction, master-mix preparation, PCR, and amplicon recovery; aliquoting of primers and probes in a zero-airflow cabinet; use of uracil dNTPs and uracil-N-glycosylase(Roche, Indianapolis, Indiana) in master mixes; and daily decontamination of all processing areas with 70% ethanol and ultraviolet light.

Laboratory investigators were blinded to clinical data and those reading culture and PCR results were blinded to one another’s interpretations. PCR data for each probe were independently classified as positive or negative for M.TB by two investigators(JAK, CH), with a consensus reading reached after joint review of discrepant results.

Smear microscopy

NTLP staff examined sputum specimens with direct Ziehl-Neelsen microscopy on the day of enrollment, according to a standard protocol.[1] Results were collected prospectively for inpatients, and by review of TB laboratory and case registries at the NTLP clinic.

Mycobacterial Culture

Frozen sputum specimens were processed at the NIH Clinical Center Mycobacteriology Laboratory using a standard NALC-NaOH procedure[16], and concentrated by centrifugation. Mycobacterial cultures were performed on Middlebrook 7H11 agar plates(Remel, Lenexa, Kansas) and in mycobacterial-growth-indicator tubes(MGIT; Beckton Dickinson, Sparks, Maryland). Acid-fast colonies were quantified, and identified using the Kinyoun stain and the Gen-Probe MTB complex Accuprobe(Gen-Probe, San Diego, California). Positive MGIT broths were examined with acid-fast and Gram stains and sub-cultured to 7H11 agar for confirmation.

To assess bias in the gold-standard culture results caused by freezing and repeat decontamination before culture at NIH, sputum from each inpatient was plated on Lowenstein-Jensen media before freezing but after standard NALC-NaOH processing at NTRL.[16] Cultures were read weekly and considered negative if no growth was identified after eight weeks.

Statistical Analysis

We used STATA 9.0(Stata Corporation, College Station, Texas) for statistical analyses, and defined significance as a two-tailed, type-I error(p-value) <0.05. We compared inpatient and outpatient characteristics and symptoms using the chi-squared test for dichotomous and categorical variables, and the Mann-Whitney rank-sum test for non-normally-distributed continuous variables. We calculated sensitivity, specificity, and positive- and negative-predictive values of sputum and oral-wash PCR assays. We used the dichotomous results of sputum mycobacterial cultures performed at NIH as the “gold-standard” outcome for both PCR specimens. We planned in advance to compare diagnostic performance by HIV status, smear status, and inpatient-versus-outpatient status using McNemar’s test for paired samples and a two-sample test of proportions for unpaired samples. Assuming a TB prevalence of 60% and a sample size of 120 patients, we calculated precision for the estimated sensitivity(88%) to be ± 7%.

Results

Participants

Of 181 outpatients enrolled in the parent study, 114 were consecutively selected(Figure 1). Six were unable to expectorate sputum; one was unable to perform oral wash. One was missing clinical data, and five were missing sputum, leaving 101 outpatients for analysis. Of 891 inpatients, 114 were eligible, and 28 were randomly selected for inclusion(Figure 1). Two were unable to expectorate sputum, leaving 26 inpatients for analysis. More patients were unable to expectorate sputum(eight) than were unable to perform oral wash(one). Forty-six percent(58/126) had HIV; in one patient, HIV status was unknown.

Figure 1.

Enrollment Diagram

Inpatients were older(median age 33 years, interquartile range(IQR) 28–42) than outpatients(median age 28 years, IQR 24–35, p=0.01), and had a higher prevalence of HIV(81% vs. 37%, p<0.001). Most HIV-infected inpatients had advanced disease(median CD4+ T-cell count 42 cells/μl, IQR 13–296). Inpatients had more fever and more severe weight loss than outpatients, but outpatients had a longer duration of symptoms. Frequencies of cough, sputum, dyspnea, and weight loss were similar(Table E1, online supplement).

Specimens

Of 127 sputums cultured at NIH, 75(59%) grew MTB complex, including 70 of 101(69%) from outpatients and five of 26(19%) from inpatients. One specimen grew Nocardia. None grew non-tuberculous mycobacteria.

PCR amplification of control sequences was successful for all sputums, and for all but two oral washes, which were inhibited at the baseline concentration and after 1:10 dilution. Baseline inhibition was more common(p<0.001) in oral washes(39/127, 31%) than in sputum(16/127, 13%). Baseline inhibition of both sputum and oral wash occurred in only three patients(one inpatient and two outpatients).

Sensitivity and Specificity of PCR on Sputum and Oral-wash Specimens

Sputum PCR was highly sensitive(99%, 95% confidence interval(CI) 93%–100%) and specific(88%, 95% CI 77%–96%) for pulmonary TB(Table 1). Oral-wash PCR was also sensitive(73%, 95% CI 62%–83%) and specific(88%, 95% CI 76%–96%) for pulmonary TB, although less sensitive than sputum PCR(difference 25%^*, 95% CI 13%–37%, p<0.001). Positive predictive values were high for sputum(92%) and oral-wash(90%) PCR, and not significantly different from one another(difference 2%, 95% CI −7%–+12%, p=0.64). In contrast, the negative predictive value of sputum PCR(98%) was significantly higher than that of oral-wash PCR(69%, difference 29%, 95% CI 17%–41%, p<0.001). Among oral-wash specimens in which baseline PCR inhibition occurred, sensitivity was significantly lower than among oral-wash specimens in which baseline PCR inhibition did not occur(52% vs. 83%, difference 31%, 95% CI 8%–53%, p=0.006).

Table 1.

Diagnostic performance of PCR tests for TB, stratified by smear and HIV status

Testa	Positive Culture (N)	Negative Culture (N)	Percent Sensitivity (95% CI)	Percent Specificity (95% CI)
All participants (n=127)	75	52
Sputum PCR			99 (93 – 100)	88 (77 – 96)
Positive	74	6
Negative	1	46
Oral-wash PCRb			73 (62 – 83)	88 (76 – 96)
Positive	55	6
Negative	20	44
Z-N sputum smear-negativec (n=37)	8	29
Sputum PCR			100 (63 – 100)	90 (73 – 98)
Positive	8	3
Negative	0	26
Oral-wash PCRb			63 (25 – 92)	85 (66 – 96)
Positive	5	4
Negative	3	23
Z-N sputum smear-positivec (n=63)	55	8
Sputum PCR			98 (90 – 100)	63 (25 – 92)
Positive	54	3
Negative	1	5
Oral-wash PCR			80 (67 – 90)	75 (35 – 97)
Positive	44	2
Negative	11	6
HIV-infectedd (n=58)	23	35
Sputum PCR			100 (85 – 100)	86 (70 – 95)
Positive	23	5
Negative	0	30
Oral-wash PCR			52 (31 – 73)	86 (70 – 95)
Positive	12	5
Negative	11	30
Testa	Positive Culture (N)	Negative Culture (N)	Percent Sensitivity (95% CI)	Percent Specificity (95% CI)
HIV-uninfectedd (n=68)	52	16
Sputum PCR			98 (90 – 100)	94 (70 – 100)
Positive	51	1
Negative	1	15
Oral-wash PCRb			83 (70 – 92)	93 (66 – 100)
Positive	43	1
Negative	9	13

Abbreviations: CI, confidence interval; PCR, polymerase chain reaction; Z-N, Ziehl-Neelsen.

^aLegend: Reference standard mycobacterial culture performed at NIH Clinical Center Microbiology Lab.

^bOral-wash PCR results inhibited for 2 patients.

^cZ-N results unavailable for 27 patients.

^dHIV result missing for 1 patient.

Among HIV-infected patients, sputum PCR sensitivity(100%, 95% CI 85%–100%) was similar(difference 2%, 95% CI −1%–9%, p=0.34) to sputum PCR sensitivity among HIV-uninfected patients(98%, 95% CI 90%–100%)(Table 1). In contrast, oral-wash PCR sensitivity was lower(difference 31%, 95% CI 8%–53%, p=0.006) among those with HIV(52%, 95% CI 31%–73%) than in those without HIV(83%, 95% CI 70%–92%). Specificities in HIV-infected and -uninfected strata were similar for sputum and oral-wash PCR.

Among patients with negative direct Ziehl-Neelsen sputum smears, sensitivity of sputum PCR was 100%(95% CI 63%–100%) and specificity was 90%(95% CI 73%–98%)(Table 1). Sensitivity of oral-wash PCR was 63%(95% CI 25%–92%) and specificity was 85%(95% CI 66%–96%). The precision of these estimates was limited by the number of smear-negative patients with positive cultures(n=8). Sensitivity and specificity estimates were similar among those with unknown sputum-smear status.

Although HIV prevalence and other characteristics varied between inpatients and outpatients(Online supplement, Table E1), there were no notable differences in the sensitivity of either PCR assay, although inpatients were few enough that some differences may have been statistically undetectable. Specificity, however, was significantly lower in inpatients for both sputum(76% for inpatients vs. 97% for outpatients, difference 21%, 95% CI 1.3%–40%, p=0.022) and oral wash(70% for inpatients vs. 100% for outpatients, difference 30%, 95% CI 10%–50%, p=0.001)(Table 2). Five of six false-positive sputum PCR results, and all six false-positive oral-wash PCR results came from inpatients.

Table 2.

Diagnostic performance of PCR tests for TB, stratified by patient population

Testa	Positive Culture (N)	Negative Culture (N)	Percent Sensitivity (95% CI)	Percent Specificity (95% CI)
Outpatients (n=101)	70	31
Sputum PCR			97 (90 – 100)	97 (83 – 100)
Positive	68	1
Negative	2	30
Oral-wash PCRb			76 (64 – 85)	100 (88 – 100)
Positive	53	0
Negative	17	30
Inpatients (n=26)	5	21
Sputum PCR			100 (48 – 100)	76 (53 – 92)
Positive	5	5
Negative	0	16
Oral-wash PCRb			60 (15 – 95)	70 (46 – 88)
Positive	3	6
Negative	2	14

Abbreviations: CI, confidence interval; PCR, polymerase chain reaction.

^aLegend: Reference standard is mycobacterial culture performed at the National Institutes of Health Clinical Center Microbiology Laboratory.

^bOral-wash PCR results inhibited for 1 outpatient and 1 inpatient.

Of 26 sputum samples cultured in both labs, six samples positive at NTRL were negative at NIH and one sample negative at NTRL was positive at NIH. To assess bias associated with using cultures of shipped, frozen sputum as the gold standard, we compared diagnostic performance of sputum PCR using two different gold standards: NTRL results on unfrozen sputum versus NIH results on previously frozen, twice-decontaminated sputum(Table E2, online supplement). Although precision was limited by the number of specimens, point estimates for sputum PCR sensitivities were similar whether measured in reference to sputum cultures done at NTRL(90%, 95% CI 56%–100%), or to those done at NIH(100%, 95% CI 48%–100%). Specificity trended higher if NTLP cultures(94%, 95% CI 70%–100%) rather than NIH cultures(76%, 95% CI 53%–92%) served as the reference. The lower bound of the 95% confidence interval of this difference suggested that NIH cultures, however imperfect, did not meaningfully overestimate specificity(difference 18%, 95% CI −3.7%–40%, p=0.14).

Discussion

In most TB-reference laboratories, specific diagnosis of pulmonary TB depends on sputum smears and cultures. While microscopic examination of sputum is generally rapid and specific, microscopy misses a significant proportion of those with TB[17]. This is especially true among those with HIV, in whom sensitivity of microscopy ranges from 36%–73%.[18] Because smear, the only available rapid diagnostic method, is insensitive, many smear-negative TB patients are not treated early enough or are not treated at all. These delays may be particularly costly in HIV-infected patients, in whom any postponement of treatment can be fatal.[10]

Because sputum smears are insensitive and labor-intensive, there has been considerable interest in using nucleic acid amplification and molecular probes for rapid diagnosis of smear-negative TB. In smear-negative specimens, commercial molecular assays are highly specific(97–99%) and have sensitivities ranging from 57–76%(Table 3).[12] Results can be obtained within four hours, even for a large number of specimens.

Table 3.

Diagnostic performance of sputum secA1 PCR test for smear-negative TB compared to commercially available sputum nucleic acid amplification assays

Test	Percent Sensitivity (95% CI)	Percent Specificity (95% CI)
BD Probe Tec ET	71 (66 – 76)	97 (96 – 97)
Cobas Amplicor MTB	64 (59 – 69)	99 (99 – 99)
Gen-Probe E-MTD	76 (70 – 80)	97 (97 – 97)
Roche Amplicor MTB	61 (57 – 65)	97 (97 – 97)
secA1	100 (63 – 100)	90 (73 – 98)

Abbreviations: CI, confidence interval; PCR, polymerase chain reaction.

Legend: BD ProbeTec ET, BD Probe Tec ET Direct Tuberculosis assay (Beckton Dickinson Diagnostics, Sparks, Maryland); Cobas Amplicor MTB, Amplicor M. tuberculosis PCR assay (Roche Molecular Systems, Branchburg, New Jersey); Gen-Probe E-MTD, amplified M. tuberculosis direct test (Gen-Probe, San Diego, California); Roche Amplicor MTB, Amplicor M. tuberculosis PCR assay (Roche Molecular Systems, Branchburg, New Jersey). Performance of commercial assays is based on pooled estimates from a previously published meta-analysis.[12]

In this study, we assessed two novel approaches. First, we used species- and genus-specific amplification assays to detect patients infected with M.TB. When sputum PCR was compared to culture, sensitivity and specificity were excellent in both smear-positive and in smear-negative patients, with sensitivity as high as existing commercial assays(Table 3). The secA1 assay also performed well in HIV-infected patients, with high positive-and negative-predictive values. Such robust performance data make this sequence a promising candidate for routine molecular diagnosis of TB in areas with a high prevalence of HIV and TB. Further testing in populations with a higher prevalence of non-tuberculous mycobacteria is needed to evaluate another possible advantage of the secA1 species- and genus-specific probes: higher specificity than smear microscopy.

This study also assessed use of oral washes for M.TB detection. In patients who cannot produce sputum, oral washes may be easier and safer to obtain than induced sputum, which can be uncomfortable and dangerous for hypoxic patients, and time-consuming and hazardous for staff. Since oral wash involves only five brief coughs, it may generate fewer infectious aerosols than sputum induction, which requires deep coughing for up-to-thirty minutes. In this study, all but one of the patients who did not expectorate sputum could perform oral wash.

SecA1 PCR of oral wash was not as sensitive as secA1 PCR of sputum. This result was expected since oral washes contain fewer secretions from the lower respiratory tract, and are diluted by saline. Also, baseline inhibition of PCR in oral washes occurred frequently and was associated with a 31% reduction in sensitivity compared to oral washes without inhibition. PCR at intermediate dilutions might improve assay sensitivity in future studies.

In this study, PCR specificity may be underestimated. The gold standard was based on a single sputum specimen that was first frozen, and then cultured months after collection.[19] While freezing does not significantly damage DNA, it may reduce viability of living organisms for culture. In addition, decontamination of inpatient sputum twice(once before each culture) may have further reduced mycobacterial viability; population-stratified analyses show that nearly all false-positive results came from these specimens. Therefore, the specificity estimates in outpatients may best reflect the true performance of these assays. Finally, a single sputum specimen may underestimate TB prevalence. Had patients provided more samples, or had information about response to TB treatment been included, more patients might have been identified as having M.TB rather than being classified as uninfected. It is even possible that secA1 PCR is more sensitive than the reference standard(mycobacterial culture), so that false-positive PCR results represented actual cases of TB.

There are several strengths to this study. First, few studies have evaluated nucleic-acid amplification for TB diagnosis in sub-Saharan Africa[5, 6, 20–22], where HIV-associated smear-negative TB is a leading cause of death.[10] Our results showing the secA1 assay performed well independently of smear or HIV status are promising for its use with sputum or oral wash in such settings, particularly if the technique is modified into a rapid, inexpensive format.

Second, we recruited participants either randomly or consecutively from clinically relevant populations living in a low-income country with high burdens of HIV and TB. Biased methods of recruitment and interpretation, including non-consecutive patient sampling and non-blinded readings, are common in studies of PCR for TB diagnosis, and have been associated with exaggerated estimates of test performance.[9, 23] In our study, interpretations of the reference and index tests were blinded and strict quality-control measures were followed for both cultures(including identification of all mycobacteria) and PCR(including testing of positive and negative controls with each PCR run), in accordance with expert recommendations for studies of TB diagnostics.[24, 25]

In summary, direct-PCR testing of oral washes or sputum holds considerable promise to become a rapid, sensitive, specific approach to identify patients with active TB, independent of HIV or smear status. As nucleic-acid amplification technology becomes more available in low-and middle-income countries, rapid molecular testing of sputum or oral washes may reduce transmission and morbidity, reducing the worldwide impact of tuberculosis.